ATI Radeon HD 2X00 (2400,2600,2900) series owners thread [Text View] - Page 23

Wo0zy

06-01-08, 12:20 PM

I've realised I know nothing about EDID. Is there a simple guide out there?

According to Moninfo my TV supports 48-62Hz and the only two resolutions mentioned are 720x480@50Hz and 1920x1080@60Hz. However, in practice the card drivers seem to realise it can use just about any resolution and refresh between the two including 800x600, 720p, 1080i etc.

Hi jong1,

I learnt everything I know from Archibael and being involved in the DTDCalc thread (it's a tool for creating custom resolutions for systems with Intel Graphics). One of DTDCalc’s features is an EDID Interpreter. Simply copy your EDID into it and click the "interpret" button and it will list all supported video and audio modes contained within.

Alternatively, http://en.wikipedia.org/wiki/EDID is a look introduction to learning EDID structures and values.

Your can get DTDCalc Here (http://www.clevertec.co.uk/productsfree.htm). Obviously only the "interpret" function will work on an ATi-based system but it saves a lot of reading. The EDID needs to contain the data only (no line numbers). In other words 8 (or 16) lines or 16 Hexadecimal values

For example

02 03 1e 77 4f 94 13 12 16 11 15 1f 05 04 03 07
02 06 10 01 23 09 07 07 65 03 0c 00 30 00 01 1d
00 bc 52 d0 1e 20 b8 28 55 40 c4 8e 21 00 00 1e
02 3a 80 d0 72 38 2d 40 10 2c 45 80 c4 8e 21 00
00 1e 01 1d 80 18 71 1c 16 20 58 2c 25 00 c4 8e
21 00 00 9e 8c 0a d0 8a 20 e0 2d 10 10 3e 96 00
c4 8e 21 00 00 18 02 3a 80 18 71 38 2d 40 58 2c
45 00 c4 8e 21 00 00 1e 00 00 00 00 00 00 00 71

Is good but

Raw EDID extension (CEA-861)
00: 02 03 1E 77 4F 94 13 12 16 11 15 1F 05 04 03 07
10: 02 06 10 01 23 09 07 07 65 03 0C 00 10 00 01 1D
20: 00 BC 52 D0 1E 20 B8 28 55 40 C4 8E 21 00 00 1E
30: 02 3A 80 D0 72 38 2D 40 10 2C 45 80 C4 8E 21 00
40: 00 1E 01 1D 80 18 71 1C 16 20 58 2C 25 00 C4 8E
50: 21 00 00 9E 8C 0A D0 8A 20 E0 2D 10 10 3E 96 00
60: C4 8E 21 00 00 18 02 3A 80 18 71 38 2D 40 58 2C
70: 45 00 C4 8E 21 00 00 1E 00 00 00 00 00 00 00 91

Won't work.

If you have a 16-line EDID (most modern TVs will), the first 8-lines contain the basic EDID information while the second 8 lines contain the Extended (Version 3) data.

Within the EDID you will find video modes defined in three ways

"Established Timings"- Codes which relate to VESA standard timing specifications.

"SVDs" (Short Video Descriptors)- A shorthand method of specifying which CEA/EIA-861B standard timings are supported

"DTD" (Detailed Timing Descriptors)- Blocks of 18 Hex values (bytes) which define all of the information regarding specific timings supported by the display.

Most displays will have at least one DTD (but not always) which is used to define the detailed timing characteristics of the displays "native" or “preferred” resolution(s).

Obviously there’s a stack more information as well but understanding the video timings and audio support parts are the most useful to us.

I'd suggest having a play with DTDCalc for a quick answer to your EDID question then reading the Wiki for a detailed description then PM me or Archibael with any specific questions. I “think” I’ve got a reasonable grasp of it now but, like I said, Archi is the real expert.

Hope this helps,

Wo0zy

PS Sorry for the long post.

jong1

06-01-08, 01:38 PM

Thanks for that Wo0zy,

I will definitely have a play later.

HT Slider

06-01-08, 02:49 PM

I've realised I know nothing about EDID. Is there a simple guide out there?

According to Moninfo my TV supports 48-62Hz and the only two resolutions mentioned are 720x480@50Hz and 1920x1080@60Hz. However, in practice the card drivers seem to realise it can use just about any resolution and refresh between the two including 800x600, 720p, 1080i etc.

In reality, unless the video card interprets your HDTV as a generic PC monitor, it actually uses HDTV formats for everything. Most likely, as well as 720x480 and 1920x1080, the driver also probably assumes it supports 1280x720@60Hz giving it 720x480, 1280x720 and 1920x1080 to work with.

When you are using 800x600 for example, it scales all output into 1280x720 and sends it to your HDTV in 720p. When using 1280x1024, it scales all output into 1920x1080 and sends it to your HDTV in 1080i (or p if your HDTV support it).

Also, if the video card decides that your HDTV provides an overscanned image, instead of scaling everything into the native 720p or 1080i/p resolutions, it scales everything into the "custom resolution" for 720p or 1080i/p instead. In other words, lets say your HDTV has 4% overscan in horizontal and 3% in vertical so you have created custom resolutions of 1232x702 and 1848x1048. The driver now is smart enough to use these custom "window" resolutions (I'm calling them a "window" because it still uses the 720p or 1080i/p full resolution to send to the card, but only the "window" in use is the custom resolution you created) for every single PC resolution.

In other words your 800x600 gets scaled into 1232x702 and then the center 1232x702 pixels are used within the full 1280x720 using 720p timing. Your HDTV still receives 720p, but it looks like 800x600 from your software's perspective. As far as you are concerned you see what looks like a perfect 800x600 displayed on your HDTV, even though the HDTV doesn't actually receive 800x600 at all.

The EDID that your HDTV sends to the video card during handshaking is what tells the driver which standard HDTV formats your HDTV supports (and if it supports true PC resolutions and timings; I'm not aware of any HDTV that does actually support true PC resolutions and timings though). By using Powerstrip to create a custom driver or creating your own .inf driver for the HDTV and then installing it, you can, to some extent, override the EDID data and allow the video card to drive the HDTV using additional resolutions and timings. These days you don't usually need to create a custom driver since CCC allows you to force most standard HDTV formats (like 1080p/50, 1080p/60, etc.), but you can still create a custom driver to work around certain limitations.

I played around with custom drivers for my Toshiba 51H83 quite a bit and always found it difficult to get multiple HDTV formats to work at the same time. In general, if you provide a range, the video card sometimes thinks it can actually use the range of timings when in reality an HDTV can only handle very specific timings. You can add a small, specific frequency range and specific resolutions and then tell the video driver to treat it as an HDTV. This worked in some situations for me (but I still couldn't get all of the non-standard formats that my HDTV supports like 540p, 960i, and 1440i to work without using Powerstrip). Powerstrip has the ability to completely override video timings so you can drive your HDTV with absolutely anything if you use it in combination with a custom HDTV driver.

jong1

06-01-08, 02:58 PM

Yeah. Good spot on 800x600. I knew that. I don't know what made me think earlier it was supported natively.

The TV seems to declare it supports 480p, 576i,720p (50/60hz), 1080i(50/60Hz), 1080p (50/60Hz) at least and I have got 1080p/48Hz working using Powerstrip. I'm just surprised hardly any of these appear in the Moninfo report. i'll be interested to see what DTDCalc manages to find!

miami_sin

06-01-08, 04:21 PM

No, that's only for PDVD Bluray/HDDVD playback. You can't do anything about accelerating older 1080p x264 files - the problem is the files break the playback standards. Ones made recently are tending to work though.

Quote:
Originally Posted by miami_sin
I found your explanation thread...
Do I have to adjust the corresponding value?

Quote:
Originally Posted by arfster
No, that's only for PDVD Bluray/HDDVD playback. You can't do anything about accelerating older 1080p x264 files - the problem is the files break the playback standards. Ones made recently are tending to work though.
Oh, OK... Didn't know that! Thanks for your quick reply though!

Then I'll go for 720p And hope newer 1080p are working!

-------------------------------------------------------------------

Any news on the above? :confused:
I'm still trying to get the 1080p stuff H/W-accelerated in Media Player Classic...
(720p is working perfectly in MPC, 1080-BluRay-Untouched are also working in Power DVD with 5% CPU)

Thanks in advance!

karrih

06-01-08, 04:33 PM

If you post the EDID you do get we can check. Also Powerstrip can probably be used to force the issue.

Wo0zy

OK, here goes

$ffffffffff ----------------------------------------------------------------
$ffffffffff Monitor EDID
$ffffffffff ----------------------------------------------------------------
$ffffffffff +00+01+02+03+04+05+06+07+08+09+0a+0b+0c+0d+0e+0f
$0c00000000 +00 00 ff ff ff ff ff ff 00 0f 30 22 91 00 00 00 00
$0c00000001 +10 32 10 01 03 80 3c 22 78 2a 03 20 a7 55 45 96 24
$0c00000002 +20 11 49 4b 1f df 00 45 59 a9 40 81 80 31 59 01 01
$0c00000003 +30 01 01 01 01 01 01 01 1d 80 18 71 1c 16 20 58 2c
$0c00000004 +40 25 00 5b 56 21 00 00 9e 8c 0a d0 8a 20 e0 2d 10
$0c00000005 +50 10 3e 96 00 5b 56 21 00 00 18 00 00 00 fc 00 43
$0c00000006 +60 41 54 2d 36 30 36 36 0a 20 20 20 20 00 00 00 fd
$0c00000007 +70 00 38 4b 1f 32 08 00 0a 20 20 20 20 20 20 01 e5
$0c00000008 +80 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c00000009 +90 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000a +a0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000b +b0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000c +c0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000d +d0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000e +e0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000f +f0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

I have RivaTuner installed for adjusting fan speed and it seems to have some resolution creation functionality. Have to look into that. I also tried to use moninfo, but it returns silly numbers and wont let me select display.

arfster

06-01-08, 04:49 PM

Any news on the above? :confused:
I'm still trying to get the 1080p stuff H/W-accelerated in Media Player Classic...

Like before: if the files aren't encoded properly, they won't work.

Wo0zy

06-01-08, 06:19 PM

In reality, unless the video card interprets your HDTV as a generic PC monitor, it actually uses HDTV formats for everything. Most likely, as well as 720x480 and 1920x1080, the driver also probably assumes it supports 1280x720@60Hz giving it 720x480, 1280x720 and 1920x1080 to work with.

When you are using 800x600 for example, it scales all output into 1280x720 and sends it to your HDTV in 720p. When using 1280x1024, it scales all output into 1920x1080 and sends it to your HDTV in 1080i (or p if your HDTV support it).

Also, if the video card decides that your HDTV provides an overscanned image, instead of scaling everything into the native 720p or 1080i/p resolutions, it scales everything into the "custom resolution" for 720p or 1080i/p instead. In other words, lets say your HDTV has 4% overscan in horizontal and 3% in vertical so you have created custom resolutions of 1232x702 and 1848x1048. The driver now is smart enough to use these custom "window" resolutions (I'm calling them a "window" because it still uses the 720p or 1080i/p full resolution to send to the card, but only the "window" in use is the custom resolution you created) for every single PC resolution.

In other words your 800x600 gets scaled into 1232x702 and then the center 1232x702 pixels are used within the full 1280x720 using 720p timing. Your HDTV still receives 720p, but it looks like 800x600 from your software's perspective. As far as you are concerned you see what looks like a perfect 800x600 displayed on your HDTV, even though the HDTV doesn't actually receive 800x600 at all.

The EDID that your HDTV sends to the video card during handshaking is what tells the driver which standard HDTV formats your HDTV supports (and if it supports true PC resolutions and timings; I'm not aware of any HDTV that does actually support true PC resolutions and timings though). By using Powerstrip to create a custom driver or creating your own .inf driver for the HDTV and then installing it, you can, to some extent, override the EDID data and allow the video card to drive the HDTV using additional resolutions and timings. These days you don't usually need to create a custom driver since CCC allows you to force most standard HDTV formats (like 1080p/50, 1080p/60, etc.), but you can still create a custom driver to work around certain limitations.

I played around with custom drivers for my Toshiba 51H83 quite a bit and always found it difficult to get multiple HDTV formats to work at the same time. In general, if you provide a range, the video card sometimes thinks it can actually use the range of timings when in reality an HDTV can only handle very specific timings. You can add a small, specific frequency range and specific resolutions and then tell the video driver to treat it as an HDTV. This worked in some situations for me (but I still couldn't get all of the non-standard formats that my HDTV supports like 540p, 960i, and 1440i to work without using Powerstrip). Powerstrip has the ability to completely override video timings so you can drive your HDTV with absolutely anything if you use it in combination with a custom HDTV driver.

I agree with your assessment regarding how HDTVs handle/scale PC resolutions and how "custom timings" are actually standard timings with a reduced number of active pixels and increased blanking interval however, don't forget that EDID version 3 data deals with HDTV timings. The SVDs define the supported EIA/CEA standard timings (including the popular 720,1080i and 1080p resolutions) and DTD's are still be used as a method of further defining a given display's preferred timing parameters for these resolutions (decode an EDID from any modern HDTV and you'll find version 3 data including both SVDs and DTDs). I don't know if ATI drivers ignore DTDs in favour of just checking SVDs and delivering standard timings but it would be a shame if that was the case. In either case, there is no assumption. The standard offerings are defined by EDID.

Powerstrip can certainly be used to override EDID parameters (that's kind of the whole point) and CCC's HDTV Support options can be used to "force" an unsupported HDTV timing if you desire but EDID still plays a major role in deciding what video timings (and audio formats) are supported and offered as standard within a driver.

Wo0zy

Wo0zy

06-01-08, 06:33 PM

OK, here goes

$ffffffffff ----------------------------------------------------------------
$ffffffffff Monitor EDID
$ffffffffff ----------------------------------------------------------------
$ffffffffff +00+01+02+03+04+05+06+07+08+09+0a+0b+0c+0d+0e+0f
$0c00000000 +00 00 ff ff ff ff ff ff 00 0f 30 22 91 00 00 00 00
$0c00000001 +10 32 10 01 03 80 3c 22 78 2a 03 20 a7 55 45 96 24
$0c00000002 +20 11 49 4b 1f df 00 45 59 a9 40 81 80 31 59 01 01
$0c00000003 +30 01 01 01 01 01 01 01 1d 80 18 71 1c 16 20 58 2c
$0c00000004 +40 25 00 5b 56 21 00 00 9e 8c 0a d0 8a 20 e0 2d 10
$0c00000005 +50 10 3e 96 00 5b 56 21 00 00 18 00 00 00 fc 00 43
$0c00000006 +60 41 54 2d 36 30 36 36 0a 20 20 20 20 00 00 00 fd
$0c00000007 +70 00 38 4b 1f 32 08 00 0a 20 20 20 20 20 20 01 e5
$0c00000008 +80 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c00000009 +90 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000a +a0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000b +b0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000c +c0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000d +d0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000e +e0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
$0c0000000f +f0 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

I have RivaTuner installed for adjusting fan speed and it seems to have some resolution creation functionality. Have to look into that. I also tried to use moninfo, but it returns silly numbers and wont let me select display.

Ok, that EDID block contains the following Established Timings
40 x 480 @ 67 Hz
640 x 480 @ 72 Hz
640 x 480 @ 75 Hz
800 x 600 @ 56 Hz
800 x 600 @ 60 Hz
800 x 600 @ 72 Hz
800 x 600 @ 75 Hz
1024 x 768 @ 87 Hz (Interlaced)
1024 x 768 @ 60 Hz
1024 x 768 @ 70 Hz
1024 x 768 @ 75 Hz
1280 x 1024 @ 75 Hz

It also contains two DTD's

01 1D 80 18 71 1C 16 20 58 2C 25 00 5B 56 21 00 00 9E (1920x1080i)
8C 0A D0 8A 20 E0 2D 10 10 3E 96 00 5B 56 21 00 00 18 (1280x720p)

However, you may have noticed that the second 8 lines are all zeros. This means the extended EDID information isn't being parsed. This second block is likely to contain either the DTDs and/or SVD's (see my previous posts) which define your 1080p50 timing.

If 1080p60 is definitely offered with the splitter then, as per HT's post, the ATI drivers are just offering you what is considered a "standard" HDTV timing but can't build on this by reading the EDID information.

Is it possible that you can temporarily disconnect the splitter and produce another report with the PC directly connected to the TV? If so and the Extended EDID does contain a DTD for 1080p50, you can use this information with Powerstrip to produce the exact timing you need. If it only offers 1080p50 in SVD then using a generic (EIA/CEA) timing with Powerstrip will work fine. In either case you may have to adjust to compensate for overscan.

Wo0zy

edavis530

06-01-08, 11:40 PM

You'll have to dload powerstrip then. Can't remember quite how, but it was pretty easy to do - you just make a new driver with 1920*1080 (or whatever) as the max possible.

How do I make my own driver? Anyone know?

I've never used powerstrip.. Also is this a free download?

HT Slider

06-02-08, 12:36 AM

I agree with your assessment regarding how HDTVs handle/scale PC resolutions and how "custom timings" are actually standard timings with a reduced number of active pixels and increased blanking interval however, don't forget that EDID version 3 data deals with HDTV timings. The SVDs define the supported EIA/CEA standard timings (including the popular 720,1080i and 1080p resolutions) and DTD's are still be used as a method of further defining a given display's preferred timing parameters for these resolutions (decode an EDID from any modern HDTV and you'll find version 3 data including both SVDs and DTDs). I don't know if ATI drivers ignore DTDs in favour of just checking SVDs and delivering standard timings but it would be a shame if that was the case. In either case, there is no assumption. The standard offerings are defined by EDID.

Powerstrip can certainly be used to override EDID parameters (that's kind of the whole point) and CCC's HDTV Support options can be used to "force" an unsupported HDTV timing if you desire but EDID still plays a major role in deciding what video timings (and audio formats) are supported and offered as standard within a driver.

Wo0zy

I guess I'm a little out of date when it comes to typical HDTV EDIDs. I didn't realize the newer HDTVs were actually reporting support of PC resolutions and higher refresh rates. I haven't really looked at EDIDs in detail for a few years and back in ~2005, all of the EDIDs I examined from HDTVs all only supported HDTV resolutions and formats (plus their single native resolution - like 1366x768). There were no 800x600's, 1024x768's or 1280x1024's in the ones I looked at back then. I'm also still under the impression that ATI only outputs HDTV formats to anything it considers an HDTV, with the specific exception to allowing 1360x768 at 60Hz. I could be wrong.

karrih, which HDTV is that EDID from?

This is even more of a reason to consider upgrading to a modern 1080p digital display (especially if it has a fast enough pixels to benefit from above 60 Hz for video games, etc.).

karrih, with your HDTV hooked up, can you select 1280x1024 at 75Hz from CCC? If it shows 1280x1024 at only 60Hz, my suspicion is it is still using 1080p for the actual output (with the current driver).

HT Slider

06-02-08, 01:00 AM

edavis530,

Powerstrip can be used for free (at least it used to), but it pops up annoying "buy me messages" unless you pay for it. It is still a great tool though (fully functional trial) and I did pay for it myself. Have a look at http://entechtaiwan.com/index.shtm for a download and additional information.

woozy,

Maybe you could take a look at my EDID and see if there is an explanation for the grey level performance I am seeing with the HDMI dongle in use (and still requiring brightness/contrast tweaks). Personally I have no idea how to interpret the color characteristics.

Here is the interpretation as well as the EDID from moninfo (as you can see my HDTV doesn't even list 720p, but it does accept and display it; as well as 540p, 960i and 1440i if I use Powerstrip):

Monitor
Windows description......... Generic PnP Monitor
Manufacturer description.... Toshiba TV
Manufacturer................ Toshiba
————————————————————————————
Plug and Play ID............ TSB0100
Serial number............... 168290
EDID data source............ Registry (stored)
————————————————————————————
Manufacture date............ 2003, ISO week 31
EDID revision............... 1.3
Display type and signal..... Digital
Sync input support.......... n/a
Screen size................. n/a
Power management............ n/a

Color characteristics
Display gamma............... 1.00
Red chromaticity............ Rx 0.640 - Ry 0.330
Green chromaticity.......... Gx 0.300 - Gy 0.600
Blue chromaticity........... Bx 0.150 - By 0.060
White point (default)....... Wx 0.313 - Wy 0.329

Timing characteristics
VESA GTF support............ Not supported
Horizontal scan range....... 31-46kHz
Vertical scan range......... 59-61Hz
Video bandwidth............. 80MHz
Extension blocks............ 1
Timing recommendation #1.... 1920x540 at 60Hz
Modeline................ "1920x540" 74.250 1920 2008 2052 2200 540 542 547 562 +hsync +vsync
Timing recommendation #2.... 1920x1080 at 30Hz
Modeline................ "1920x1080" 74.250 1920 2008 2052 2200 1080 1084 1094 1124 interlace +hsync +vsync
Timing recommendation #3.... 720x480 at 60Hz
Modeline................ "720x480" 27.000 720 736 798 858 480 489 495 525 -hsync -vsync

Standard timings supported
640 x 480 at 60Hz - IBM VGA
720 x 480 at 60Hz - Toshiba
1920 x 540 at 60Hz - Toshiba
1920 x 1080 at 30Hz - Toshiba

Raw EDID base
00: 00 FF FF FF FF FF FF 00 52 62 00 01 62 91 02 00
10: 1F 0D 01 03 80 00 00 00 0E EE 91 A3 54 4C 99 26
20: 0F 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
30: 01 01 01 01 01 01 01 1D 80 18 71 1C 16 20 58 2C
40: 25 00 00 00 00 00 00 9E 8C 0A D0 8A 20 E0 2D 10
50: 10 3E 96 00 00 00 00 00 00 18 00 00 00 FC 00 54
60: 6F 73 68 69 62 61 20 54 56 0A 20 20 00 00 00 FD
70: 00 3B 3D 1F 2E 08 00 0A 20 20 20 20 20 20 01 E2

mariner888

06-02-08, 02:05 AM

I suspect the CEA extension block in EDID is not read into the registry by the latest Catalyst drivers. Not sure if it's by design or a bug.

No problem with the Forceware drivers.

karrih

06-02-08, 02:09 AM

Is it possible that you can temporarily disconnect the splitter and produce another report with the PC directly connected to the TV? If so and the Extended EDID does contain a DTD for 1080p50, you can use this information with Powerstrip to produce the exact timing you need. If it only offers 1080p50 in SVD then using a generic (EIA/CEA) timing with Powerstrip will work fine. In either case you may have to adjust to compensate for overscan.

Wo0zy

I guess I'll try direct connection too. Some kind of cut and paste EDID operation for 1080p50 would indeed be ideal.

karrih

06-02-08, 02:34 AM

karrih, which HDTV is that EDID from?

karrih, with your HDTV hooked up, can you select 1280x1024 at 75Hz from CCC? If it shows 1280x1024 at only 60Hz, my suspicion is it is still using 1080p for the actual output (with the current driver).

When I took the report I had Panasonic TH42-PX80 plasma and Mitsubishi HC1100 projector connected using HDMI splitter. Mits was in standby, though. I don't know how splitter handles EDID from those two displays, but we might find out soon...

Using 1080p when choosing 1280x1024 sounds odd, but then I haven't quite figured out either why choosing 1080i25 in CCC gets reported as 1080i50 by my pj.

madshi

06-02-08, 03:30 AM

however, don't forget that EDID version 3 data deals with HDTV timings. The SVDs define the supported EIA/CEA standard timings (including the popular 720,1080i and 1080p resolutions) and DTD's are still be used as a method of further defining a given display's preferred timing parameters for these resolutions (decode an EDID from any modern HDTV and you'll find version 3 data including both SVDs and DTDs). I don't know if ATI drivers ignore DTDs in favour of just checking SVDs and delivering standard timings but it would be a shame if that was the case. In either case, there is no assumption. The standard offerings are defined by EDID.
I know how older ATI drivers behaved. Haven't checked out the latest drivers yet, though. With older drivers ATI did pay attention to the DTDs. However, IMO it's actually ***BAD*** to do that. Why? Because the pixel clock in DTDs is just plain wrong for 24Hz and 60Hz. What we need for HTPC playback is 24/1.001 and 60/1.001 and the DTDs don't have enough decimal digits for that purpose. So a standalone Blu-Ray player actually ignores at least the pixel clock in the DTD and uses EIA/CEA standard timings instead. ATI didn't do that in older drivers and that is BAD. Furthermore ATI interpreted the SVDs in such a way that 24Hz is 24.000Hz and 60Hz is 60.000Hz. That is wrong, too. It should be 23.976Hz and 59.940Hz.

But as I said that's all with older drivers. Have yet to check the latest drivers where according to the release notes some changes were made to this stuff.

jong1

06-02-08, 08:49 AM

@Wo0zy,

Do you have a suggested tool for collecting the EDID data to go into DTDCalc?

Thanks

karrih

06-02-08, 10:38 AM

Is it possible that you can temporarily disconnect the splitter and produce another report with the PC directly connected to the TV?

Turns out that if I do this with HDMI connection, at least RivaTuner reports it can not read EDID.

Wo0zy

06-02-08, 10:45 AM

@Wo0zy,

Do you have a suggested tool for collecting the EDID data to go into DTDCalc?

Thanks

With Intel Drivers the built-in diagnostic report contains the EDID (very helpful) but for other vendors I usually use Moninfo (http://www.entechtaiwan.com/util/moninfo.shtm)

I think you can also extract EDID information from within Powerstrip (not sure about that one. Will have to check).

Wo0zy

Wo0zy

06-02-08, 10:46 AM

Turns out that if I do this with HDMI connection, at least RivaTuner reports it can not read EDID.

Hmmm. Try using Moninfo. I just included the download URL in my previous post.

Wo0zy

nicksan

06-02-08, 11:54 AM

Hello,

Been reading up on this thread. Took me a while and I can't deny the fact that I am indeed pretty confused about all this!

Last week, on a whim, I decided that I wanted Blue-ray playback for my Home Theater that features a Sanyo PLV-Z4 720p Projector, Sony SD DVD Player, and an antiquated P4 Northwood 2.8 PC. (Asus P4P-800, 1.5GB DDR, Creative Audigy SE)

So rather than buying a PS3 or building a new PC from scratch I decided to buy Sony's BD-ROM and the Sapphire Radeon 2600XT AGP 8X card. I also have the full version of PowerDVD 8 ultra with most recent update. (Also made sure I saved the older codec since I am aware that the latest update disables HA.)

I installed the latest ATI Catalyst hotfix (8.5) and that's the only driver I installed. This machine never had an ATI card installed so I guess it's like a fresh install in this sense. I read about the Omega drivers but then I saw the 8.5 hotfix so decided to give that a go instead. I did apply the registry hack on the guide however.

I am running audio to spdif, output via Audigy SE and into my older Pioneer Elite receiver. It's Dolby Digital/DTS 5.1 via spdif.

I bought Casino Royale Blue-ray and that actually plays back fine in PowerDVD with HA enabled. The CPU typicall shows 40%-45% with spikes that bring it up to 50%-55%...definitely not more than 60%. No stuttering at all. Mind you, this is all in 720p. (Video card resolution set to standard 720p 1280 x 720...which is the native resolution on my Projector)

So being that this was so exciting, I decided to play the demo Blue-Ray DVD that came with the Sony BD-ROM. (Has trailers for Spiderman 3, Casino Royale, Surf's up, etc...) For whatever reason this one stutters like mad. 100% CPU despite the fact that HA is enabled.

So in PowerDVD I checked the "Information" tab and compare between this and the Casino Royale BR DVD that works well and they look pretty much the same EXCEPT...the audio encoding.

Casino Royale = Dolby Digital
Sony BD-ROM Sample BR DVD = MLP

I know MLP is more "advanced" in that it is lossless but could this actually affect the HA and and cause stuttering? In PowerDVD since I send the audio to spdif, and because the receiver doesn't do MLP, I choose to mix it down to either DD or DTS. This has no effect. It stutters like crazy!

I borrowed a BR DVD from my co-worker, Oceans 13, and will see how that plays, however I am concerned about all this. The last thing I want to do is to upgrade my machine to a Core2Duo...my wife would kill me!:eek:

So gathering from what I have described, is there anything I am doing wrong? Again, Casino Royale BR DVD plays fine. So it's not a complete wash. There is something on that Sony BD-Rom sample BR DVD that is causing the CPU to spike up and cause stutter and the only difference I can tell is, again, the MLP audio vs. regular DD 5.1 on Casino Royale.

Any help would be greatly appreciated.

jong1

06-02-08, 11:57 AM

karrih

06-02-08, 12:40 PM

I also have the full version of PowerDVD 8 ultra with most recent update. (Also made sure I saved the older codec since I am aware that the latest update disables HA.)

Casino Royale = Dolby Digital
Sony BD-ROM Sample BR DVD = MLP

I know MLP is more "advanced" in that it is lossless but could this actually affect the HA and and cause stuttering?

Any help would be greatly appreciated.

I am also running underpowered setup with HA and until PDVD Ultra 7.3 build 4102 I was not able to use Dolby TrueHD without major stutter. Now I can use TrueHD with 720p but not with 1080p due to the HA problems you mention. The secondary audio stream usually worked for me, though.

nicksan

06-02-08, 12:51 PM

I take it Dolby TrueHD is the same thing as MLP?

I also do have PDVD 7.3 installed and it stutters. I am not sure if I have 4102 though...I think it's lower than that so perhaps it's worth a try.

You would figure though that with PDVD 8 Ultra being a newer version they would have ironed out the kinks.

Perhaps it's specific to my system?

When you say secondary audio stream, you mean when I down-mix the MLP to DD 5.1 and send it to spdif?

Perhaps this problem is specific to the Sampler BR DVD in question. I am tempted to upgrade my rig, however I guess the prudent thing to do is to just buy/rent some BR DVD's and start watching them.

The thing is though, I am wondering if this problem is with BR DVD's authored more recently and that I won't have problems with BR DVD's authored in 2007 but the newer release will have this problem?

In these BR DVD's is there an audio option screen like regular DVD's? On my Casino Royale BR DVD I cound't get the menu to come up and had to screw around to get the movie to start. Is that a common problem?

Maybe I should just invest in a G35 motherboard + Core2Duo (or quad)....

I am also running underpowered setup with HA and until PDVD Ultra 7.3 build 4102 I was not able to use Dolby TrueHD without major stutter. Now I can use TrueHD with 720p but not with 1080p due to the HA problems you mention. The secondary audio stream usually worked for me, though.

karrih

06-02-08, 01:03 PM

Hmmm. Try using Moninfo. I just included the download URL in my previous post.

Wo0zy

Here it is, but doesn't look too promising to me. EDID source registry, eh?

Monitor
Windows description......... Generic Television
Manufacturer................ PNP
————————————————————————————
Plug and Play ID............ PNP09FE
Serial number............... n/a
EDID data source............ Registry (stored)
————————————————————————————
Manufacture date............ 1990
EDID revision............... 1.3
Display type and signal..... Analog 0.700,0.300 (1.0V p-p)
Sync input support.......... n/a
Screen size................. n/a
Power management............ Standby, Suspend, Active off/sleep

Color characteristics
Display gamma............... 1.00
Red chromaticity............ Rx 0.000 - Ry 0.000
Green chromaticity.......... Gx 0.000 - Gy 0.000
Blue chromaticity........... Bx 0.000 - By 0.000
White point (default)....... Wx 0.000 - Wy 0.000

Timing characteristics
VESA GTF support............ Not supported
Horizontal scan range....... 30-48kHz (est.)
Vertical scan range......... 48-60Hz (est.)
Video bandwidth............. 64MHz (est.)
Extension blocks............ n/a

Standard timings supported
640 x 480 at 60Hz - IBM VGA
800 x 600 at 60Hz - VESA
1024 x 768 at 60Hz - VESA

Raw EDID base
00: 00 FF FF FF FF FF FF 00 41 D0 FE 09 00 00 00 00
10: 00 00 01 03 00 00 00 00 F0 00 00 00 00 00 00 00
20: 00 00 00 21 08 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 D1

karrih

06-02-08, 01:10 PM

I take it Dolby TrueHD is the same thing as MLP?

When you say secondary audio stream, you mean when I down-mix the MLP to DD 5.1 and send it to spdif?

In these BR DVD's is there an audio option screen like regular DVD's? On my Casino Royale BR DVD I cound't get the menu to come up and had to screw around to get the movie to start. Is that a common problem?

Right now I can't remember if TrueHD uses MLP, but it could well be.

In addition to TrueHD there is often option to use another audio format, for example PCM or DD 5.1 (not downmixed) and those are lighter for processor. You can change audio streams using buttons in PDVD when there is no audio menu at hand.

nicksan

06-02-08, 01:24 PM

I looked it up and it looks to be the case regarding TrueHD=MLP/lossless.

I am aware of the PDVD audio stuff...what I do is select spdif as the output and it comes up with option whether to use Primary Audio or something else. (I forget)

I guess using Primary just passes the MLP straight up and my receiver only does 2.1. There is another option to down-mix to DD 5.1 / DTS. I choose this option and my receiver does 5.1.

However I am not sure when down-mixing, it is still processing MLP as well and the down mix is secondary?

Got a headache now!:eek::)

Right now I can't remember if TrueHD uses MLP, but it could well be.

In addition to TrueHD there is often option to use another audio format, for example PCM or DD 5.1 (not downmixed) and those are lighter for processor. You can change audio streams using buttons in PDVD when there is no audio menu at hand.

nicksan

06-02-08, 02:04 PM

Also I am wondering maybe there is a problem with the fact that there is no menu on the sampler BR DVD that allows me to choose the audio format. Seems to force using MLP...maybe that's what is causing the 100% spike and stutter?

In Casino Royale, even though I cannot for the life of me reach or see the main menu, it is selecting DD 5.1...maybe it's selected by default...or that it doesn't have TrueHD being that it's an "older" release?

Who knows...but it's kind of difficult to figure out a pattern unless I view more BR DVD's I suppose. I guess I shouldn't let the sampler BR DVD that came with the Sony BD-Rom drive to scare me into upgrading my PC!

For those BR DVD that have TrueHD, do they typically have "separate" DD5.1 tracks on them as well? Perhaps if that is the case, I have nothing to be concerned about. Not sure how all that works...

As for my h.264 file, it seems to run fine, whether I am using PDVD or MPC w/CoreAVC codecs. So for simplicity I may just use PDVD as my only player.

Curious to see how the others like WinDVD and Arcsoft performs. Too bad WinDVD9 trial cannot do BR. Stupid...just plain stupid. How am I supposed to evaluate it for BR?:confused:

Right now I can't remember if TrueHD uses MLP, but it could well be.

In addition to TrueHD there is often option to use another audio format, for example PCM or DD 5.1 (not downmixed) and those are lighter for processor. You can change audio streams using buttons in PDVD when there is no audio menu at hand.

Wo0zy

06-02-08, 02:56 PM

I am using an ATI 3850.

I already had looked at Moninfo, but that only has the 2 timings in it. Posted the raw data into DTDCalc and I get the same.

I am guessing I am not getting the extended stuff, if it exists. I know it's a bit off-topic but any tips?

Hmm. I tried Moninfo for the first time on my ATI system and I get the same problem. It does the same thing with Intel Graphics (not that it's a problem as we can get a full EDID from the driver diags report) however, the last time I used it with an NVIDIA card it captured a complete EDID including the extended (1.3) data. Very odd.

Not sure what to suggest. The Phoenix EDID editor does the same thing and I can't get the Viewsonic editor to work properly under Vista. :(

I'll ask Archibael if he's aware of anything.

AFAIK you can't export EDID from within Powerstrip but if you right-click on the taskbar icon and select Options>Monitor Information, it lists supported timings. You could try this and see if if offers you any additional timings.

Sorry

Wo0zy

jong1

06-02-08, 03:02 PM

AFAIK you can't export EDID from within Powerstrip but if you right-click on the taskbar icon and select Options>Monitor Information, it lists supported timings. You could try this and see if if offers you any additional timings.No, sadly this gives the same minimal info. Maybe that IS all that is there and it guesses the rest. I really do not know. Not that I have a problem, but I am intrigued now to know what gems my TV really does divulge!

HT Slider

06-02-08, 03:32 PM

No, sadly this gives the same minimal info. Maybe that IS all that is there and it guesses the rest. I really do not know. Not that I have a problem, but I am intrigued now to know what gems my TV really does divulge!

Which HDTV do you have again?

It is very strange to see only standard PC resolutions in the EDID and no HDTV resolutions.

All of the HDTV EDID's I've personally extracted and looked at only have HDTV resolutions (as well as often a standard 640x480).

I suspect that for some reason the EDID isn't being read properly and the numbers moninfo is displaying are some sort of default.

HT Slider

06-02-08, 03:33 PM

Hmm. I tried Moninfo for the first time on my ATI system and I get the same problem. It does the same thing with Intel Graphics (not that it's a problem as we can get a full EDID from the driver diags report) however, the last time I used it with an NVIDIA card it captured a complete EDID including the extended (1.3) data. Very odd.

WoOzy, did you notice my post a page back? (http://www.avsforum.com/avs-vb/showthread.php?p=13993473#post13993473)

I wondered if you might have some insight into the grey level performance I am seeing with my HDTV.

SugoE

06-02-08, 05:30 PM

@Wo0zy,

Do you have a suggested tool for collecting the EDID data to go into DTDCalc?

Thanks
I used RivaTuner to read out EDID, maybe it's more accurate then Moninfo.

jong1

06-02-08, 05:40 PM

Which HDTV do you have again?

It is very strange to see only standard PC resolutions in the EDID and no HDTV resolutions.

All of the HDTV EDID's I've personally extracted and looked at only have HDTV resolutions (as well as often a standard 640x480).

I suspect that for some reason the EDID isn't being read properly and the numbers moninfo is displaying are some sort of default.It's a Sony 46X2000 but I didn't say it only reads PC resolutions.

According to Moninfo my TV supports 48-62Hz and the only two resolutions mentioned are 720x480@50Hz and 1920x1080@60Hz. However, in practice the card drivers seem to realise it can use just about any resolution and refresh between the two including 800x600, 720p, 1080i etc.

As I said before I don't know what was going through my mind when I mentioned 800x600 there. Clearly nonsense!

In practice it seems to support 480p/i, 576p/i, 720p and 1080p/i @48-60Hz and devices seem to recognise that, but I am mystified how!

I used RivaTuner to read out EDID, maybe it's more accurate then Moninfo.No. tried that one too! But thanks.

jong1

06-02-08, 05:45 PM

WoOzy, did you notice my post a page back? (http://www.avsforum.com/avs-vb/showthread.php?p=13993473#post13993473)

I wondered if you might have some insight into the grey level performance I am seeing with my HDTV.Maybe I am missing something, but as per my earlier post this seems clear. Unless the dongle is "in action", switching the card to "HDMI mode", then the card outputs RGB @PC-levels. It seems clear from what is appearing in your CCC that, despite the dongle, the driver is recognising, correctly, that it is in fact connected to a DVI display, that it therefore cannot use YCbCr (which always uses video levels) and so it is treating it like a PC monitor.

Is there a way that this does not fit with the facts?

Wo0zy

06-02-08, 06:03 PM

I used RivaTuner to read out EDID, maybe it's more accurate then Moninfo.

Nice one SugoE. The RivaTuner EDID report does capture all information (including 1.3 extended data).

Jong1, once you create the report it'll need a bit of notepad editing before you can run it through DTDCalc....but it does work.

Wo0zy

Edit: Just noticed that you said it didn't for you?? Once I selected "Monitor EDID snapshot" and "Monitor EDID details", I got the following

$ffffffffff ----------------------------------------------------------------
$ffffffffff Monitor EDID
$ffffffffff ----------------------------------------------------------------
$ffffffffff +00+01+02+03+04+05+06+07+08+09+0a+0b+0c+0d+0e+0f
$0c00000000 +00 00 ff ff ff ff ff ff 00 52 62 04 01 00 00 00 00
$0c00000001 +10 00 10 01 03 80 c4 8e 78 0a ee 91 a3 54 4c 99 26
$0c00000002 +20 0f 50 54 00 00 00 01 01 01 01 01 01 01 01 01 01
$0c00000003 +30 01 01 01 01 01 01 01 1d 80 d0 72 1c 16 20 10 2c
$0c00000004 +40 25 80 c4 8e 21 00 00 9e 8c 0a d0 90 20 40 31 20
$0c00000005 +50 0c 40 55 00 c4 8e 21 00 00 18 00 00 00 fc 00 54
$0c00000006 +60 4f 53 48 49 42 41 2d 54 56 0a 20 20 00 00 00 fd
$0c00000007 +70 00 31 3d 0f 44 0f 00 0a 20 20 20 20 20 20 01 eb
$0c00000008 +80 02 03 1e 77 4f 94 13 12 16 11 15 1f 05 04 03 07
$0c00000009 +90 02 06 10 01 23 09 07 07 65 03 0c 00 10 00 01 1d
$0c0000000a +a0 00 bc 52 d0 1e 20 b8 28 55 40 c4 8e 21 00 00 1e
$0c0000000b +b0 02 3a 80 d0 72 38 2d 40 10 2c 45 80 c4 8e 21 00
$0c0000000c +c0 00 1e 01 1d 80 18 71 1c 16 20 58 2c 25 00 c4 8e
$0c0000000d +d0 21 00 00 9e 8c 0a d0 8a 20 e0 2d 10 10 3e 96 00
$0c0000000e +e0 c4 8e 21 00 00 18 02 3a 80 18 71 38 2d 40 58 2c
$0c0000000f +f0 45 00 c4 8e 21 00 00 1e 00 00 00 00 00 00 00 91
$ffffffffff ----------------------------------------------------------------
$ffffffffff Monitor EDID details
$ffffffffff ----------------------------------------------------------------
$0d00000000 EDID version : 1.3
$0d00000001 Manufacturer ID : TSB (Toshiba)
$0d00000002 Model ID : 0104 (TOSHIBA-TV)
$0d00000003 Manufactured on : 2006, week 0
$0d00000004 Serial number : 00000000
$0d00000005 Input signal : digital
$0d00000006 Screen size : 196 x 142 cm
$0d00000007 Horizontal frequency : 15-68KHz
$0d00000008 Vertical frequency : 49-61Hz
$0d00000009 Bandwidth : 150MHz

The EDID block matches exactly with the report that the Intel Diagnostic report produces

jong1

06-02-08, 06:14 PM

I'll look at it again tomorrow. I may have been too quick to judge!

Wo0zy

06-02-08, 06:34 PM

WoOzy, did you notice my post a page back? (http://www.avsforum.com/avs-vb/showthread.php?p=13993473#post13993473)

I wondered if you might have some insight into the grey level performance I am seeing with my HDTV.

Sorry HT, I missed that post.

I think Jong1 is correct. Your connection isn't end-to-end HDMI therefore the connection is treated as RGB/PC rather than YCbCr/video levels. I use my main system without the dongle and with my TV calibrated (as best I can) for RGB levels and using only the "UseBT601CSC=1" reg tweak, I'm fairly happy that all my levels (desktop, SD and HD) are consistent. At least they are acceptable to my eyes. All video looks good and photos are detailed enough for casual viewing (I don't use this system for editing anyway). Obviously, switching to a YCbCr input (such as my DVB-T tuner or Sky receiver) reveals a rubbish washed out, over bright image but this doesn't bother me as I use the HTPC for just about eveything anyway :)

Regarding your EDID report, as per other posts, it only seems to contain the first block (no extended data). Perhaps you could try RivaTuner as SugoE suggests and see if it's possible to capture the Version 1.3 extended block? It's likely that the 720p timing is in there as either an SVD or DTD.

Wo0zy

Dee_NA

06-02-08, 07:27 PM

Wo0zy

06-02-08, 07:58 PM

Here it is, but doesn't look too promising to me. EDID source registry, eh?

Monitor
Windows description......... Generic Television
Manufacturer................ PNP
————————————————————————————
Plug and Play ID............ PNP09FE
Serial number............... n/a
EDID data source............ Registry (stored)
————————————————————————————
Manufacture date............ 1990
EDID revision............... 1.3
Display type and signal..... Analog 0.700,0.300 (1.0V p-p)
Sync input support.......... n/a
Screen size................. n/a
Power management............ Standby, Suspend, Active off/sleep

Color characteristics
Display gamma............... 1.00
Red chromaticity............ Rx 0.000 - Ry 0.000
Green chromaticity.......... Gx 0.000 - Gy 0.000
Blue chromaticity........... Bx 0.000 - By 0.000
White point (default)....... Wx 0.000 - Wy 0.000

Timing characteristics
VESA GTF support............ Not supported
Horizontal scan range....... 30-48kHz (est.)
Vertical scan range......... 48-60Hz (est.)
Video bandwidth............. 64MHz (est.)
Extension blocks............ n/a

Standard timings supported
640 x 480 at 60Hz - IBM VGA
800 x 600 at 60Hz - VESA
1024 x 768 at 60Hz - VESA

Raw EDID base
00: 00 FF FF FF FF FF FF 00 41 D0 FE 09 00 00 00 00
10: 00 00 01 03 00 00 00 00 F0 00 00 00 00 00 00 00
20: 00 00 00 21 08 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 D1

Very odd! This is when directly connected to your TV via HDM with no other display connected?

It's almost like your display reports no EDID and it's picked up a generic entry of some sort :confused:

Registry (stored) is normal but the the lack of data and the fact the connection is reported as Analog is very strange.

Perhaps try again with RivaTuner as SugoE discovered.

Wo0zy

arfster

06-02-08, 07:59 PM

@ Arfster - If you were me, would you choose the HD 2400 PRO or the HD 3650? Just figured I'd ask since i'm tryin to decide which one is better since I only use the cards for movies especially when I bought em both at differ times of course.

3650 definitely. It's more powerful, as the 3650 is a 2600 equivalent, and the 2400 cards have all sorts of daft issues (can't do decent deinterlacing for a start). ATI also have a bit of a habit of giving up on old cards too, so those bugs will never be fixed.

Best of all, prices have dropped so there's not much difference between them now :-)

karrih

06-03-08, 01:10 AM

Perhaps try again with RivaTuner as SugoE discovered.

Wo0zy

That was my first attempt. RivaTuner reports it can not read EDID. Perhaps ATI offers all HDTV resolutions when the connection is recognized as HDMI and there is no EDID? If my splitter takes my projectors EDID and passes it as its own then ATI thinks that it knows what is available and no more offers all HDTV resolutions as standard.

Wo0zy

06-03-08, 07:12 AM

That was my first attempt. RivaTuner reports it can not read EDID. Perhaps ATI offers all HDTV resolutions when the connection is recognized as HDMI and there is no EDID? If my splitter takes my projectors EDID and passes it as its own then ATI thinks that it knows what is available and no more offers all HDTV resolutions as standard.

I guess it's possible but it's very unusual for an HDTV not to have even a basic EDID, especially a known brand like Panasonic.

Didn't you produce the first report from RivaTuner with the splitter still connected?

It might be worth clearing out the monitor entries from your registry and trying a RivaTuner report again without the splitter.

Wo0zy

jong1

06-03-08, 07:52 AM

Wo0zy

06-03-08, 12:23 PM

For me Rivatuner also reports no extended data. It shows a table covering the extended area but all after line 7 is blank.

I cannot believe this, as having read a bit more about all this I now know that the beginning of the extended area is where an HDMI device is required to declare that it is HDMI, to distinguish itself from DVI using the "HDMI Vendor-Specific Data Block":

http://www.hdmi.org/download/HDMI_Specification_1.1.pdf

It may be to do with a "splitter". My video is routed through my AV amp (Denon 4306). But it is clear from the above document that the repeater is required to pass through the EDID data so the end device can operate correctly. Maybe it is just all these tools do not read the data correctly when a repeater is in the loop, although for compatibility reasons I cannot see why.

This is very strange. It must be the tools.

If the extended data block wasn't passed back with "Repeaters" in the chain then Intel based HTPC's would not send HDMI audio (Intel are very anal about EDID ;)) and (other than the issue of malformed SAD's in some Denon AVRs) we know they do. Regardless, the extended block is where the supported audio formats are stored and it would be ridiculous to think that AVR's can't report that information :rolleyes:

Don't suppose you're able to hook up the HTPC directly to the TV just for a quick test?

Wo0zy

HT Slider

06-03-08, 12:28 PM

Maybe I am missing something, but as per my earlier post this seems clear. Unless the dongle is "in action", switching the card to "HDMI mode", then the card outputs RGB @PC-levels. It seems clear from what is appearing in your CCC that, despite the dongle, the driver is recognising, correctly, that it is in fact connected to a DVI display, that it therefore cannot use YCbCr (which always uses video levels) and so it is treating it like a PC monitor.

Is there a way that this does not fit with the facts?

While that might be a logical explanation, the fact that when I remove the HDMI dongle and use DVI straight to DVI, the grey level problems become twice as bad, suggests otherwise. With the dongle and default settings I can see grey levels from 16-235 (bitmap displayed) and without the dongle and default settings I can see roughly grey levels from 30-220 (same bitmap).

Since the dongle does get rid of one level of expansion when compared to DVI direct, there must be some other explanation, other than the ATI card doesn't recognize a difference between HDMI dongle vs no HDMI dongle when an actual DVI port is used at least.

One explanation for example could be that my HDTV is simply out of calibration and can only display 30-220. From a brightness/contrast adjustment sense it definitely isn't out of calibration though (I can crank up the brightness and turn down the contrast until the display is totally washed out, yet it never shows any additional grey levels). I suppose it could somehow be out of calibration from strictly a digital perspective (able to only display 30-220), but this does not really make sense to me. How can the DVI port digitally be that far out? On top of that, why don't I see the same problem when an Nvidia card is used (it has many different problems, but desktop by default can visualize 16-235 grey levels without any adjustment to the Nvidia control panel when using DVI-DVI; as expected. Only ATI cards do this extra level of expansion on everything).

I think it is far more likely that the EDID is somehow confusing the ATI card into performing an extra level of expansion as a final step on output, but I don't know enough about the EDID to figure out if this is possible. Thus the reason I was asking the question.

HT Slider

06-03-08, 12:35 PM

Sorry HT, I missed that post.

I think Jong1 is correct. Your connection isn't end-to-end HDMI therefore the connection is treated as RGB/PC rather than YCbCr/video levels. I use my main system without the dongle and with my TV calibrated (as best I can) for RGB levels and using only the "UseBT601CSC=1" reg tweak, I'm fairly happy that all my levels (desktop, SD and HD) are consistent. At least they are acceptable to my eyes. All video looks good and photos are detailed enough for casual viewing (I don't use this system for editing anyway). Obviously, switching to a YCbCr input (such as my DVB-T tuner or Sky receiver) reveals a rubbish washed out, over bright image but this doesn't bother me as I use the HTPC for just about eveything anyway :)

Regarding your EDID report, as per other posts, it only seems to contain the first block (no extended data). Perhaps you could try RivaTuner as SugoE suggests and see if it's possible to capture the Version 1.3 extended block? It's likely that the 720p timing is in there as either an SVD or DTD.

Wo0zy

I didn't include the extended data because it was all 0's.

I do believe they really are 0's too because I've read the EDID using many different methods over the years with this particular HDTV and the extended EDID has always come up as 0's, even when reading from the HDTV directly.

Wo0zy

06-03-08, 12:37 PM

jong1,

Try EDID Viewer

http://www.softpedia.com/get/System/...D-Viewer.shtml

It doesn't show SAD's (although it does show that my TV supports "Basic Audio") but is able to show EDID EXTENSION : EIA/CEA-861 video data including SVDs and DTD's.

It's interesting that this tool and Moninfo read the EDID from the registry. Perhaps clearing out the entries will force a redetection?

Wo0zy

Wo0zy

06-03-08, 01:05 PM

I didn't include the extended data because it was all 0's.

I do believe they really are 0's too because I've read the EDID using many different methods over the years with this particular HDTV and the extended EDID has always come up as 0's, even when reading from the HDTV directly.

Yeah. that's more than possible. Toshiba used to be quite lazy with their EDIDs.

Judging from mine (a Tosh circ 2006) they finally sorted in out :).

Having said that, in Version 1.3 byte 126 specifies the number of extension blocks and yours suggests that there is one. As per my post to jong1, might be worth seeing what EDID Viewer comes up with just out of curiosity.

Wo0zy

jong1

06-03-08, 02:09 PM

While that might be a logical explanation, the fact that when I remove the HDMI dongle and use DVI straight to DVI, the grey level problems become twice as bad, suggests otherwise. With the dongle and default settings I can see grey levels from 16-235 (bitmap displayed) and without the dongle and default settings I can see roughly grey levels from 30-220 (same bitmap).

Since the dongle does get rid of one level of expansion when compared to DVI direct, there must be some other explanation, other than the ATI card doesn't recognize a difference between HDMI dongle vs no HDMI dongle when an actual DVI port is used at least.

One explanation for example could be that my HDTV is simply out of calibration and can only display 30-220. From a brightness/contrast adjustment sense it definitely isn't out of calibration though (I can crank up the brightness and turn down the contrast until the display is totally washed out, yet it never shows any additional grey levels). I suppose it could somehow be out of calibration from strictly a digital perspective (able to only display 30-220), but this does not really make sense to me. How can the DVI port digitally be that far out? On top of that, why don't I see the same problem when an Nvidia card is used (it has many different problems, but desktop by default can visualize 16-235 grey levels without any adjustment to the Nvidia control panel when using DVI-DVI; as expected. Only ATI cards do this extra level of expansion on everything).

I think it is far more likely that the EDID is somehow confusing the ATI card into performing an extra level of expansion as a final step on output, but I don't know enough about the EDID to figure out if this is possible. Thus the reason I was asking the question.Hmm. That does sound odd. When you say "bitmap" I assume you mean a photo - not a frozen video frame - with all 255 levels contained within it?

From what I have read over the last day I do not think the EDID should have any control over this. Not to say ATI could not have messed things up somehow though!

nicksan

06-03-08, 02:37 PM

Sorry to interrupt this highly technical conversation going on here...:)

Anyways, I did end up borrowing Ocean's 13 BR and it played back fine. Wasn't too impressed with the production of the movie, however that's another story! Again in the Information page I see it's using Dolby Digital 5.1, not MLP...so I guess it wasn't a great test. I need to get my hands on a title with TrueHD to see if it messes everything up again.

Also, once again, I was unable to get the main menu up at all. What's up with that?

Does PDVD figure out the audio streams on the BR DVD? What I mean by this is selecting the DD5.1 AC-3 track via the BR menu (which I can't reach at the moment...) the same as selecting the audio track in PDVD itself?

I am just concerned that b/c I can never seem to get at the menu for these BR titles to select the audio track, which for me will always be AC-3 DD5.1 or DTS since I have an older receiver, that if I encounter a title with TrueHD as the default track, then it will mess things up. That's what is happening with the Sony Sample BR disc. Stuttering like crazy...audio = MLP...and no way to choose DD5.1 b/c there isn't a section in the menu to select the audio track. Maybe it doesn't even have an AC-3 track?

Right now I can't remember if TrueHD uses MLP, but it could well be.

In addition to TrueHD there is often option to use another audio format, for example PCM or DD 5.1 (not downmixed) and those are lighter for processor. You can change audio streams using buttons in PDVD when there is no audio menu at hand.

HT Slider

06-03-08, 03:29 PM

Hmm. That does sound odd. When you say "bitmap" I assume you mean a photo - not a frozen video frame - with all 255 levels contained within it?

From what I have read over the last day I do not think the EDID should have any control over this. Not to say ATI could not have messed things up somehow though!

I'm using the calibration bitmap (bmp) that I posted about a page back with the links to various calibration sources. It contains a ramp of all 256 levels.

EDID does contain color information. I'm not certain what, but I wouldn't be surprised if it contains supported grey levels as well as RGB and/or YCbCr support.

Where did you read that EDID has no control over color?

Is there a separate DVI/HDMI handshaking protocol that negotiates color?

Something does....

jong1

06-03-08, 05:52 PM

Wo0zy

06-03-08, 06:28 PM

Look honestly, I don't claim to be an expert on EDID at all. Just trying to help troubleshoot. But have a look at that HDMI document I posted a link to a little while back.

It says that YCbCr must be "limited range" (16-235) and RGB may be either. It is not negotiated, just stated. It also says that the colorspace to be used can be inferred by the resolution or explicitly stated in an optional AVI Infoframe broadcast along with a video stream, but even in that info frame there is no info about the levels that should be used.

Of course that is not the DVI spec and your TV is DVI, so things could be different there, but still I'd be surprised.

You're one fast reader jong1 ;)

That's a truly mind-blowing document to digest!! Having said that, I have to agree. My sanitised understanding suggests that EDID is unlikely to be the source of HT's issue.

Firstly, he has no extension data available to comment on (despite the 1.1 data block claiming there is one). Secondly, the 1.1 data is almost identical in all important ways to mine and I don't experience the same problems. However, my display IS HDMI.

HT, page 115 onwards of the HDMI specification document is very interesting with regard to HDMI>DVI Compatibility. Well worth a read.

Wo0zy

bobbyrr

06-03-08, 07:06 PM

Anyone know why I might be getting unsmooth playback of a 1080p blue ray video on my hd2600xt?

Using powerdvd 8 v1662, vista 32bit sp1.

Possibly related, ticking and unticking use ati avivo hardware acceleration does not change the cpu usage of poweredvd, it stays around 38-45% on my core 2 duo 2.66ghz cpu.

catalyst 8.5s and latest AVIVO installed.

archibael

06-03-08, 08:56 PM

I wrote up a long message with footnotes and everything (it was impressive), then my crappy laptop crashed and I lost it. I'll summarize:

Per the HDMI 1.3a spec, an HDMI source is not allowed to send YCbCr to an HDMI sink at anything other than video levels (16/235), and can only send full range (0/255) on RGB if the resolution is 640x480.

HOWEVER, if a source doesn't see the second EDID block-- which I think it's clear from this thread that the ATI drivers do not-- it has to assume the sink is a DVI device, and it must send RGB. Unfortunately, I don't have a copy of the DVI 1.0 spec, so I can't tell what conditions (if any) require video levels. Presumably, the source would be sending full range by default.

The EDID does have some bits allocated for color space, chroma, and gamma info in the first block, but it is my experience those are ignored for the most part-- and in this case, I'm not seeing anything in there which would affect the situation.

madshi

06-04-08, 02:44 AM

HOWEVER, if a source doesn't see the second EDID block-- which I think it's clear from this thread that the ATI drivers do not
Incorrect. The ATI drivers do see the 2nd block, at least those drivers that I'm using (7.11 IIRC). I can adjust the EDID in my Lumagen Radiance video processor. So I can say that for sure.

jong1

06-04-08, 07:40 AM

@HT Slider

What are you using to display your .bmp file? Is there any chance this could be doing one level of expansion (16-235 -> 0-255)? With the driver at all default settings, have you tried taking a snapshot of your desktop with the .bmp displayed and using photoshop or other tool to check the levels?

I could imagine that if played back in some video player that it might erroneously do one level of expansion in the same way it would with video. This would be sufficient to get what you are observing above since your PC would then be losing the 0-15 data due to expansion and your TV would then be clipping the new 0-15 data (16-31 in the original) as you have explained it does.

Otherwise I would just repeat myself and suggest a complete clean uninstall and re-install of drivers and some new testing of HD & SD video and .bmp files at default, without dongle or registry tweaks using AVIVO only for video. I would using MPC-HC for video just to provide a solid easy to compare baseline.

archibael

06-04-08, 11:59 AM

Incorrect. The ATI drivers do see the 2nd block, at least those drivers that I'm using (7.11 IIRC). I can adjust the EDID in my Lumagen Radiance video processor. So I can say that for sure.

I stand corrected. I haven't been following the thread until it was pointed out to me by Wo0zy and I had gotten the impression nobody's EDID-read tools were working.

Wo0zy

06-04-08, 12:18 PM

I stand corrected. I haven't been following the thread until it was pointed out to me by Wo0zy and I had gotten the impression nobody's EDID-read tools were working.

It seems some are and some aren't. HT's EDID report suggests there's an extension block but it shows as all zeros. jong1's definitely should have an extension block but that shows as all zeros as well. I can see my extension block using just about any tool.

Weird situation :confused:

You confirmed that it's unlikely to be EDID that's causing HT's expansion problems which ticks another possibility off the list.

Thanks for taking a look.

Wo0zy

jong1

06-04-08, 01:00 PM

HT Slider

06-04-08, 01:20 PM

You confirmed that it's unlikely to be EDID that's causing HT's expansion problems which ticks another possibility off the list.

I'm not sure anything archibael explained confirms that EDID isn't involved - at least I didn't read it that way.

Keep in mind:

The ATI card does display within CCC that my HDTV is hooked up using DVI, regardless of if the HDMI dongle is used or not (but I get double expansion if no dongle is used). The double expansion could be a bug in the driver perhaps and single expansion when the HDMI dongle is used, could be because the card ultimately does realize that DVI is the connection at the HDTV somehow. The question I have here is how does the ATI card know if the HDVT's final connection is HDMI or DVI? Is this in the EDID or somewhere else? Based on what I've read, my understanding is that the end connection being HDMI vs DVI shouldn't have any effect. The only thing the video card knows is that an HDMI cable is attached to it and a specific EDID report is received from the HDTV.

My HDTV does provides, within the EDID, a specific color space, chroma and gamma details. Most devices don't seem to contain this information. Moninfo reports the following from the EDID (as posted earlier):

Color characteristics
Display gamma............... 1.00
Red chromaticity............ Rx 0.640 - Ry 0.330
Green chromaticity.......... Gx 0.300 - Gy 0.600
Blue chromaticity........... Bx 0.150 - By 0.060
White point (default)....... Wx 0.313 - Wy 0.329

All I know for certain right now:

There appears to be an extra level of expansion going on somewhere.
The EDID contains some specific information regarding colorspace, chroma and gamma (that I don't know how to interpret).
For some reason, even though the HDMI dongle is used, CCC reports the connection as DVI (noting that the actual, final connection to the HDTV is indeed DVI).
When the HDMI dongle is NOT used, my HDTV experiences a double expansion (~30-220 visible).
To answer jong1's question earlier, I am using simply the desktop or Vista picture viewer to view the color ramp. Taking a screen capture of the entire desktop and viewing it shows exactly the same color ramp proving that the same 0-255 is retained in a screen capture. Again, by default, using the HDMI dongle I see 16-235 on my HDTV and without the dongle I see roughly 30-220. Note that a "typical" display seems to experience one less level of expansion (typically no HDMI dongle 16-235 visible; HDMI dongle used 0-255 visible)
When the HDMI dongle is used, I can get correct grey levels by cranking up the brightness in CCC to +31 and turning down the contrast to 73%. While this does work, I expect it reduces the number of grey levels available for image processing.

At this point I still don't know enough, but I do know that people have successfully re-written the EDID with my HDTV. There are also methods (at least there used to be) to overwrite the EDID within the registry and change the way the video driver interprets the display. I'd like to find a way to make my ATI card drive my HDTV in a "normal" fashion (an Nvidia card can, so why can't an ATI card).

I might be looking in totally the wrong direction, but I still suspect the specific EDID my HDTV provides might be involved.

archibael

06-04-08, 01:21 PM

HT Slider

06-04-08, 01:28 PM

Reading through the HDMI specification (thanks jong1 and woozy) I found the following:

8.3.3 DVI/HDMI Device Discrimination
In order to determine if a sink is an HDMI device, an HDMI Source shall check the E-EDID for the
presence of an HDMI Vendor Specific Data Block within the first CEA EDID Timing Extension.
Any device with an HDMI VSDB of any valid length, containing the IEEE Registration Identifier of
0x000C03, shall be treated as an HDMI device.
Any device with an E-EDID that does not contain an HDMI VSDB

It appears that the EDID is what tells the video card if an HDMI or DVI device is in use. My current guess is this most likely is where ATI video cards are getting confused and sending regular RGB, even when the HDMI dongle is used. The bizarre feature is how they do a double expansion when the HDMI dongle is not used.

archibael

06-04-08, 01:37 PM

I'm not sure anything archibael explained confirms that EDID isn't involved - at least I didn't read it that way.
The question I have here is how does the ATI card know if the HDVT's final connection is HDMI or DVI? Is this in the EDID or somewhere else?

EDID second block has an HDMI vendor specific block in it if the final connection is HDMI, otherwise it's DVI. That's how it's supposed to make the choice.

Based on what I've read, my understanding is that the end connection being HDMI vs DVI shouldn't have any effect. The only thing the video card knows is that an HDMI cable is attached to it and a specific EDID report is received from the HDTV.

It knows the dongle is attached, which may put the drivers into some interesting mode (actually, it definitely puts the drivers into a different mode, as the dongle activates audio), but it has no clue about the cabling. You can have a chain of various DVI and HDMI cables with converters and the card (nor the end device) will not be able to tell the difference.

My HDTV does provides, within the EDID, a specific color space, chroma and gamma details. Most devices don't seem to contain this information.

Really? It's been my experience that most reported EDIDs do have that information.

There appears to be an extra level of expansion going on somewhere.
The EDID contains some specific information regarding colorspace, chroma and gamma (that I don't know how to interpret).
For some reason, even though the HDMI dongle is used, CCC reports the connection as DVI (noting that the actual, final connection to the HDTV is indeed DVI).
When the HDMI dongle is NOT used, my HDTV experiences a double expansion (~30-220 visible).

Yeah, this sounds like a bug.

New Driver Guy: "Hey, what's this about "fix expansion for DVI color quantization?"
Old Driver Guy: "Oh, yeah, when you have DVI you need to modify the quantization by like 15 on both ends.
New Driver Guy: "Okay. Let's see... [typing] If DVI, then RGB(max) = RGB(max) - 15; if DVI, then RGB (min) = RGB (min) + 15. There, that ought to do it."

I might be looking in totally the wrong direction, but I still suspect the specific EDID my HDTV provides might be involved.

It is possible, I'm just not seeing anything which jumps out at me. It's also possible that there's something in the registry for that particular monitor which got hosed and it lingers beyond dongle removal. When you first hooked it up, was it with the dongle or not? Perhaps finding all the registry entries for that monitor, deleting, them, and then trying again without the dongle would fix things. [shrug] Sorry, I don't have too many useful suggestions on this one.

HT Slider

06-04-08, 01:39 PM

Other possibility is that they are taking a shortcut and merely looking to see if the first block states the existence of the second (extension) block, and then just assuming from there that it's HDMI. The second-to-last byte in the first block will be a "01" if there is a second block, or a "00" if not.

Here is the full EDID extracted with Moninfo:

Raw EDID base
00: 00 FF FF FF FF FF FF 00 52 62 00 01 62 91 02 00
10: 1F 0D 01 03 80 00 00 00 0E EE 91 A3 54 4C 99 26
20: 0F 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
30: 01 01 01 01 01 01 01 1D 80 18 71 1C 16 20 58 2C
40: 25 00 00 00 00 00 00 9E 8C 0A D0 8A 20 E0 2D 10
50: 10 3E 96 00 00 00 00 00 00 18 00 00 00 FC 00 54
60: 6F 73 68 69 62 61 20 54 56 0A 20 20 00 00 00 FD
70: 00 3B 3D 1F 2E 08 00 0A 20 20 20 20 20 20 01 E2

Raw EDID extension (reserved)
00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

The second to last block is 01, but the extension is all zeros.

When the specification states "HDMI Source shall check the E-EDID for the
presence of an HDMI Vendor Specific Data Block within the first CEA EDID Timing Extension", what does that really mean?

My HDTV's E-EDID is all zero's, so I assume this means the EDID is telling the source device that is is talking to a DVI device (which also aligns with what CCC reports).

archibael

06-04-08, 01:45 PM

From the EDID Wiki:

A Vendor Specific Data Block (if any) contains as its first three bytes the vendor's IEEE 24-bit registration number, LSB first. It is usually followed by a three byte source physical address, LSB first. The source physical address provides the CEC physical address for upstream CEC devices. The remainder of the Vendor Specific Data Block is the "data payload",which can be anything the vendor considers worthy of inclusion in this EDID extension block.

Basically, in the second EDID block there is an optional sub-block of data called the VSDB. If included, it can have anything the "vendor" wants to put in there. The HDMI spec says that if you want to be recognized as a valid HDMI sink device, you must have some specific HDMI "vendor" data in the second block.

HT Slider

06-04-08, 02:00 PM

I missed your last post when I posted mine...

Really? It's been my experience that most reported EDIDs do have that information.

Perhaps I shouldn't have said don't contain it. My numbers just look different from what I've seen elsewhere. Do these numbers look typical in your (much greater) experience?

Yeah, this sounds like a bug.

New Driver Guy: "Hey, what's this about "fix expansion for DVI color quantization?"
Old Driver Guy: "Oh, yeah, when you have DVI you need to modify the quantization by like 15 on both ends.
New Driver Guy: "Okay. Let's see... [typing] If DVI, then RGB(max) = RGB(max) - 15; if DVI, then RGB (min) = RGB (min) + 15. There, that ought to do it."

I think you are bang on with this statement.

Most likely there is a bug in the ATI driver where it knows it is outputting to an HDTV and the HDTV reports a DVI connection (this would be relatively rare I would expect). Depending on which "output mode" the video card is in (by using or not using the ATI HDMI dongle), it either gets the grey levels partially confused (dongle in use) or totally confused (dongle not in use).

It is possible, I'm just not seeing anything which jumps out at me. It's also possible that there's something in the registry for that particular monitor which got hosed and it lingers beyond dongle removal. When you first hooked it up, was it with the dongle or not? Perhaps finding all the registry entries for that monitor, deleting, them, and then trying again without the dongle would fix things. [shrug] Sorry, I don't have too many useful suggestions on this one.

I guess my thinking wasn't that is EDID is actually wrong, but that the EDID is what is triggering the issue with ATI drivers. Since it is the EDID that tells the video card DVI is the final connection to an HDTV, it is most likely the EDID that triggers the bug.

If my guess is correct and there are very few individuals with HDTVs using actual DVI connections, this is likely one of those bugs that will never get fixed.

One work around for me would probably be to figure out how to reflash a new EDID to my HDTV. If I can make it appear to be using HDMI, this issue will likely go away.

Now the question becomes how easily can I reflash a new EDID (while still being able to go back to the original one later).

Maybe I should open a support request with ATI also about this bug, but it is a very complex issue and I'm still not 100% certain how it all works.

New thoughts...

We know that when the HDMI dongle is used, I end up with 16-235 visible (calibration ramp picture).
We know that when the HDMI dongle is not used, I end up with 30-220 visible (same ramp picture).
We know that my HDTV does report correctly that DVI is used (at the HDTV itself).
We also know that when an Nvidia card is used, without any HDMI dongle, that 16-235 is visible.
Essentially this means that as long as I use the ATI dongle and an HDMI to DVI cable with an ATI card that it behaves the same as an Nvidia card.

New questions:

Does this mean that there is a bug in both ATI and Nvidia drivers or is my HDTV somehow telling the video card through its EDID that it can actually display RGB (0-255) when it can only display YCbCr (16-235)?
Does the "sink" telling the "source" that DVI is used, automatically mean the sink is expected to output using RGB (0-255), or is there something else in the EDID that is supposed to tell the sink to send YcBcR (16-235)?

EDIT: I missed your very last post prior to submitting this again...

HT Slider

06-04-08, 02:08 PM

Basically, in the second EDID block there is an optional sub-block of data called the VSDB. If included, it can have anything the "vendor" wants to put in there. The HDMI spec says that if you want to be recognized as a valid HDMI sink device, you must have some specific HDMI "vendor" data in the second block.

This part makes sense. My HDTV is correctly using the EDID to tell the source that it has a DVI connection. It does have DVI connection, even though it can only display YCbCr (16-235). It cannot handle RGB (0-255).

Does the fact that it reports as DVI automatically tell the source to send it RGB or is there something else in the EDID that defines if it wants YCbCr or RGB?

Perhaps it is whatever within the EDID that tells the video card to send YCbCr that the video card is ignoring (or the EDID is missing)?

The double expansion issue does indeed seem to be a bug in the ATI driver, but as long as an HDMI dongle is used it seems to go away.

jong1

06-04-08, 02:11 PM

I looked in the registry and found the EDID for my monitor. The full extended EDID is there. I have no idea why all the tools I have tried seem not to display it.

HT I suggest you have a look in HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Enum\DISPLAY

There will probably be a lot there but look for the one with the right "DeviceDesc". Mine was @ HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Enum\DISPLAY\DON 0004\5&96259de&3&11335588&06&00

Export the "Device Parameters" key as text. Then open the text file and tidy it up (a bit fiddly).

In case anyone is interested mine in the end was:

00 ff ff ff ff ff ff 00 11 ee 04 00 01 01 01 01
00 00 01 03 80 46 28 78 0a ee 91 a3 54 4c 99 26
0f 50 54 00 00 00 01 01 01 01 01 01 01 01 01 01
01 01 01 01 01 01 02 3a 80 d0 72 38 2d 40 10 2c
45 80 df a4 21 00 00 1e 8c 0a d0 8a 20 e0 2d 10
10 3e 96 00 df a4 21 00 00 18 00 00 00 fc 00 44
45 4e 4f 4e 2d 41 56 41 4d 50 0a 20 00 00 00 fd
00 30 3e 0e 46 0f 00 0a 20 20 20 20 20 20 01 3f
02 03 31 76 51 14 03 04 12 13 05 01 07 16 9f 10
15 11 06 0f 1d 1e 23 0f 1f 07 23 09 7f 07 23 3d
1f c0 23 15 1f 51 83 5f 00 00 66 03 0c 00 21 00
80 01 1d 00 72 51 d0 1e 20 6e 28 55 00 df a4 21
00 00 1e 8c 0a d0 90 20 40 31 20 0c 40 55 00 df
a4 21 00 00 18 01 1d 00 bc 52 d0 1e 20 b8 28 55
40 df a4 21 00 00 1e 01 1d 80 18 71 1c 16 20 58
2c 25 00 df a4 21 00 00 9e 00 00 00 00 00 00 1d

Interpreted by DTD Calculator as:

No Established Timings Found
Block 1 DTD Found (See List)
Block 1 DTD Found (See List)
Display does not support underscan
Display supports basic audio
Display supports YCbCr 4:4:4
Display supports YCbCr 4:2:2
6 native formats in DTDs
Short Video Descriptor:
Resolution Supported - 1920x1080i @ 50Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 720x480p @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 1280x720p @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 720x576p @ 50Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 1280x720p @ 50Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 1920x1080i @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 640x480p @ 59.94/60Hz 4:3 Non-Native
Short Video Descriptor:
Resolution Supported - 720x480i @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 720x576i @ 50Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 1920x1080p @ 50Hz 16:9 Native
Short Video Descriptor:
Resolution Supported - 1920x1080p @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 720x576i @ 50Hz 4:3 Non-Native
Short Video Descriptor:
Resolution Supported - 720x576p @ 50Hz 4:3 Non-Native
Short Video Descriptor:
Resolution Supported - 720x480i @ 59.94/60Hz 4:3 Non-Native
Short Video Descriptor:
Resolution Supported - 1440x480p @ 59.94/60Hz 16:9 Non-Native
Short Video Descriptor:
Resolution Supported - 1440x576p @ 50Hz 4:3 Non-Native
Short Video Descriptor:
Resolution Supported - 1440x576 @ 50Hz 16:9 Non-Native
Short Audio Descriptor:
Audio Format - LPCM
8 Channel Sound
Sampling Frequencies Supported :96kHz, 88kHz, 48kHz, 44kHz, 32kHz
Bit Depth :24 bit, 20 bit, 16 bit
Short Audio Descriptor:
Audio Format - LPCM
2 Channel Sound
Sampling Frequencies Supported :192kHz, 176kHz, 96kHz, 88kHz, 48kHz, 44kHz, 32kHz
Bit Depth :24 bit, 20 bit, 16 bit
Short Audio Descriptor:
Audio Format - DTS
6 Channel Sound
Sampling Frequencies Supported :96kHz, 88kHz, 48kHz, 44kHz, 32kHz
Maximum supported bitrate :1536
Short Audio Descriptor:
Audio Format - AC-3
6 Channel Sound
Sampling Frequencies Supported :96kHz, 88kHz, 48kHz, 44kHz, 32kHz
Maximum supported bitrate :648
Speaker Allocation Block :
Speakers Present :
Rear Left Centre / Rear Right Centre
Block 2 DTD Found (See List)
Block 2 DTD Found (See List)
Block 2 DTD Found (See List)
Block 2 DTD Found (See List)

It seems to have both the TV info and additional, mainly audio, info added by my AVR.

Jon

jong1

06-04-08, 02:16 PM

Wo0zy

06-04-08, 02:26 PM

archibael

06-04-08, 02:40 PM

This part makes sense. My HDTV is correctly using the EDID to tell the source that it has a DVI connection. It does have DVI connection, even though it can only display YCbCr (16-235). It cannot handle RGB (0-255).

Does the fact that it reports as DVI automatically tell the source to send it RGB or is there something else in the EDID that defines if it wants YCbCr or RGB?

Perhaps it is whatever within the EDID that tells the video card to send YCbCr that the video card is ignoring (or the EDID is missing)?

The double expansion issue does indeed seem to be a bug in the ATI driver, but as long as an HDMI dongle is used it seems to go away.

I don't think there's anything in the EDID which tells the source to send RGB vs YCbCr other than the fact of whether it's DVI or not. If DVI, send RGB; if HDMI send either (at the discretion of the source), but must be 16-255. Though if this is something with your TV and no one else is encountering it, obviously it starts to point more at the EDID again.

As for colorspace parameters in the EDID, I'll give you some random examples of what I've seen:

Monitor
Windows description......... HannStar Monitor
Manufacturer description.... HW191A
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.643 - Ry 0.325
Green chromaticity.......... Gx 0.295 - Gy 0.616
Blue chromaticity........... Bx 0.143 - By 0.081
White point (default)....... Wx 0.310 - Wy 0.330

Monitor
Windows description......... Generic PnP Monitor
Manufacturer description.... 202M
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.640 - Ry 0.340
Green chromaticity.......... Gx 0.295 - Gy 0.610
Blue chromaticity........... Bx 0.145 - By 0.070
White point (default)....... Wx 0.313 - Wy 0.329

Monitor
Windows description......... ViewSonic Monitor
Manufacturer description.... VX2235wm-3
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.646 - Ry 0.339
Green chromaticity.......... Gx 0.290 - Gy 0.603
Blue chromaticity........... Bx 0.145 - By 0.065
White point (default)....... Wx 0.313 - Wy 0.329

Monitor
Windows description......... ViewSonic Monitor
Manufacturer description.... VX2235wm
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.640 - Ry 0.333
Green chromaticity.......... Gx 0.285 - Gy 0.602
Blue chromaticity........... Bx 0.152 - By 0.074
White point (default)....... Wx 0.313 - Wy 0.329

Monitor
Manufacturer description.... SAMSUNG
Manufacturer................ Samsung
Color characteristics
Display gamma............... 2,40
Red chromaticity............ Rx 0,632 - Ry 0,357
Green chromaticity.......... Gx 0,289 - Gy 0,596
Blue chromaticity........... Bx 0,143 - By 0,085
White point (default)....... Wx 0,280 - Wy 0,290

Monitor
Windows description......... Sony Monitor
Manufacturer description.... SONY TV
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.637 - Ry 0.321
Green chromaticity.......... Gx 0.289 - Gy 0.601
Blue chromaticity........... Bx 0.146 - By 0.081
White point (default)....... Wx 0.285 - Wy 0.294

Monitor Windows description......... Plug and Play Monitor
Manufacturer description.... DELL 2007WFP
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.640 - Ry 0.330
Green chromaticity.......... Gx 0.300 - Gy 0.600
Blue chromaticity........... Bx 0.150 - By 0.060
White point (default)....... Wx 0.313 - Wy 0.329

Monitor
Windows description......... SyncMaster 225MW(Digital)
Manufacturer description.... SyncMaster
Color characteristics
Display gamma............... 2,20
Red chromaticity............ Rx 0,639 - Ry 0,333
Green chromaticity.......... Gx 0,289 - Gy 0,597
Blue chromaticity........... Bx 0,153 - By 0,082
White point (default)....... Wx 0,313 - Wy 0,329

Monitor
Windows description......... Plug und Play-Monitor
Manufacturer description.... SANYO UK4
Color characteristics
Display gamma............... 2,30
Red chromaticity............ Rx 0,640 - Ry 0,330
Green chromaticity.......... Gx 0,290 - Gy 0,600
Blue chromaticity........... Bx 0,150 - By 0,060
White point (default)....... Wx 0,280 - Wy 0,290

archibael

06-04-08, 02:42 PM

Sweet :)

DTDCalc doesn't interpret ALL supported audio formats yet ( we're still working on the "codes" for the compressed HD formats) but it's still useful information.

I figured out what each is and updated the Wiki, but as for the details... without the 861D spec, I have no idea whether the new HD formats follow the same conventions as the older ones.

@Archibael,

Thanks again for taking the time to get involved.

Wo0zy

No problem. Sorry if I'm not much help.

jong1

06-04-08, 03:01 PM

Sweet :)

DTDCalc doesn't interpret ALL supported audio formats yet ( we're still working on the "codes" for the compressed HD formats) but it's still useful information.

Double-clicking on the DTD's extracted to the RH box will give you all the timing data. It's designed to convert to "modeline" values but you should still be able to work out the important information from the "calculation" tab.That's OK. My amp support 8-channel LPCM, but not HD bitstreams.

I did look at the DTDs. they still confuse me a bit.

- There are not DTDs for all the short descriptions by any means.

- There is a DTD for 1080i which show a very odd estimated refresh rate of something like 60.04Hz (not sure, but over 60hz anyway).

- There is no 1080p/60 DTD, but it certainly accepts it and it works perfectly.

- I don't understand the whole "native"/"non-native" thing. I thought it was just about the resolution of the panel but not according to the short descriptions.

Any thoughts?

I can only think the reason why none of the tools manage to extract the extended data is because they are getting confused by all the additional AVR data in there and dropping everything as a consequence.

archibael

06-04-08, 03:23 PM

That's OK. My amp support 8-channel LPCM, but not HD bitstreams.

I did look at the DTDs. they still confuse me a bit.

- There are not DTDs for all the short descriptions by any means.

That's correct. It is assumed that a mode with a SVD (Short Video Descriptor) uses the generic timings for that mode. DTDs would be redundant, as the source should know how to transmit general 720x480p, for instance.

- There is a DTD for 1080i which show a very odd estimated refresh rate of something like 60.04Hz (not sure, but over 60hz anyway).

Yep. The DTD notation, being integer based (except for the pixel clock) will often result in refresh rates which are slightly off. Generally speaking, the display should be able to cope with this (they're required by the EIA/CEA spec to allow for a certain percentage of pixel clock flex), but there can sometimes be problems.

- There is no 1080p/60 DTD, but it certainly accepts it and it works perfectly.

Yep. It's supported in SVD, so the source should know how to send it.

- I don't understand the whole "native"/"non-native" thing. I thought it was just about the resolution of the panel but not according to the short descriptions.

"Native" is supposed to be the resolution your TV/monitor is best able to display. Whether it is the actual resolution of the TV or whether it is just what the TV manufacturers have decided is the best scaling algorithm on their input chips, it's what a source device should nominally choose for you by default.

HT Slider

06-04-08, 03:45 PM

So in PowerDVD I checked the "Information" tab and compare between this and the Casino Royale BR DVD that works well and they look pretty much the same EXCEPT...the audio encoding.

Casino Royale = Dolby Digital
Sony BD-ROM Sample BR DVD = MLP

I know MLP is more "advanced" in that it is lossless but could this actually affect the HA and and cause stuttering? In PowerDVD since I send the audio to spdif, and because the receiver doesn't do MLP, I choose to mix it down to either DD or DTS. This has no effect. It stutters like crazy!

I borrowed a BR DVD from my co-worker, Oceans 13, and will see how that plays, however I am concerned about all this. The last thing I want to do is to upgrade my machine to a Core2Duo...my wife would kill me!:eek:

So gathering from what I have described, is there anything I am doing wrong? Again, Casino Royale BR DVD plays fine. So it's not a complete wash. There is something on that Sony BD-Rom sample BR DVD that is causing the CPU to spike up and cause stutter and the only difference I can tell is, again, the MLP audio vs. regular DD 5.1 on Casino Royale.

Any help would be greatly appreciated.

I don't think you are doing anything wrong, other than possibly expecting AGP cards to truly work as advertised and drop the CPU utilization to ultra low levels, especially when playing modern, high end, 7.1 channel audio tracks, especially if you want to transcode them into DD5.1 or DTS (using the CPU).

The reality is AGP cards are notorious for producing either no hardware acceleration (not your situation) or in general higher CPU utilization than PCIe cards (of course PCIe cards tend to also live in PCs with more CPU power).

The other (more significant in your case) reality is PowerDVD Ultra uses a lot of CPU power to decode higher end audio tracks.

More than likely you'll need to rely on selecting slightly lower end audio tracks, ideally DTS or DD5.1 and configure PowerDVD to send them directly to your audio hardware for decoding there. Although I don't know the actual number, I would guess that at least 95% of all Blu-ray movies available today contain either DTS or DD5.1 sound tracks. PowerDVD Ultra tends to automatically select high level audio track so you'll probably need to manually select DTS or DD5.1 when finding a disk that doesn't play properly (there may be a way within the options to auto select DD5.1 or DTS also; I haven't looked).

Keep in mind that DTS and DD5.1 still produces very good 5.1 audio. Unless you have either an HDMI audio solution supporting 7.1 channels or an extremely high end audio card, DTS and DD5.1 are typically the best audio tracks to use anyway. Note that not many PC HDMI devices currently are capable of 7.1 channel audio either. The ATI HD 2x00 and 3xx0 cards do not support anything higher than DTS or DD5.1, so the best audio option is to select DTS or DD5.1 anyway, even if you did have a more powerful CPU.

The other option of course is to upgrade your hardware. If you were to do any upgrades, I would highly recommend moving to a PCIe motherboard and PCIe HD video card as a key part of the upgrade.

EDIT: I had a look at my PowerDVD Ultra and I couldn't see any way to configure it to automatically select DTS or DD5.1 audio tracks. When playing a disk, often there is a setup menu where you can select DTS or DD5.1. Alternatively you can use the mouse to right click and select the DTS or DD5.1 audio track from the list of different audio tracks on that particular Blu-ray.

HT Slider

06-04-08, 04:30 PM

The HDMI spec seems clear - if DVI send RGB .

If your TV is really a DVI device that only accepts YCbCr this could certainly be leading to confusion!

I think though you are probably saying your TV clips to video levels (16-235) rather than that it must be YCbCr encoded. Are you?

I have scoured the specifications everywhere I can for my HDTV (Toshiba 51H83). From what I can tell it accepts both RGB and YCbCr through DVI, but from experience I can say with reasonable confidence that it only displays levels between 16-235.

My understanding is if I was to hook up any true video device using DVI to my TV that it would work fine (output 16-235). My only source to confirm this is talking with others with the same HDTV who are using DVI with their video equipment. How does an STB or digital video recorder know to send 16-235 instead of 0-255? Why doesn't the ATI card send 16-235 since it does recognize the TV as an HDTV.

Archibael, could you explain how a regular STB or other video device knows to send 16-235, while both Nvidia and ATI (when the HDMI dongle is also used at least) sends 0-255.

As far as why ATI sends double expanded video to my HDTV when the HDMI dongle is not used, I can only assume this must be a bug. I started to wonder if I was mistaken so I pulled out the cable and double checked straight DVI. I have to crank the CCC brightness all the way up to about +75 and contrast way down to about 35% to get compliant video levels, but if the HDMI dongle IS used, I "only" need +31 brightness and 73% contrast.

Ultimately all of this suggests that if I was to reflash a slightly different EDID to my HDTV that I could probably convince the video card to send YCbCr. As long as the HDTV does actually support this, this would probably fix everything.

archibael

06-04-08, 04:36 PM

Archibael, could you explain how a regular STB or other video device knows to send 16-235, while both Nvidia and ATI (when the HDMI dongle is also used at least) sends 0-255.

No idea. You've reached the limit of my knowledge. My assumption is that unless instructed otherwise, STBs send 16-235 (so-called "video" levels), and that it is only on a specific override that they will send 0-255, but it's possible there are the equivalent of the old STB DVD player "black level setting" option on many STBs, either in a menu or in a service mode. For HDMI end-devices, in fact, 16-235 appears mandatory.

HT Slider

06-04-08, 04:41 PM

I don't think there's anything in the EDID which tells the source to send RGB vs YCbCr other than the fact of whether it's DVI or not. If DVI, send RGB; if HDMI send either (at the discretion of the source), but must be 16-255. Though if this is something with your TV and no one else is encountering it, obviously it starts to point more at the EDID again.

I don't know how many HDTVs actually have DVI ports on them, but my guess would be that all HDTVs that do would experience the same problem.

As for colorspace parameters in the EDID, I'll give you some random examples of what I've seen:

Monitor Windows description......... Plug and Play Monitor
Manufacturer description.... DELL 2007WFP
Color characteristics
Display gamma............... 2.20
Red chromaticity............ Rx 0.640 - Ry 0.330
Green chromaticity.......... Gx 0.300 - Gy 0.600
Blue chromaticity........... Bx 0.150 - By 0.060
White point (default)....... Wx 0.313 - Wy 0.329

It is interesting that my numbers are 100% identical to the Dell 2007WFP.

I had wondered if the color characteristics were used to control if RGB 0-255 or RGB 16-235 was used. Seeing a Dell monitor having identical values to my HDTV, I now assume they have nothing to do with grey levels at all.

After researching, reading, and learning a lot from the 3 of you, am I correct to assume that there is nothing in the HDMI/DVI specification anywhere that tells a source that a sink (display) wants RGB 0-255 or RGB 16-235?

I now understand that RGB vs YCbCr is controlled by having an EDID defining the sink as DVI or HDMI, but what tells the source to send RGB 0-255 or RGB 16-235?

Is this something that was totally overlooked when the specifications were being written?

I have also been told that within the display output options for a Sony PS3 that there are 3 options RGB 0-255, RGB 16-235 and YCbCr 16-235. Sony has apparently made the decision for some reason to allow manual control over this.

archibael

06-04-08, 04:47 PM

Nothing I've seen. There might be something in the DVI 1.0 spec, but I don't have it handy.

HT Slider

06-04-08, 04:51 PM

No idea. You've reached the limit of my knowledge. My assumption is that unless instructed otherwise, STBs send 16-235 (so-called "video" levels), and that it is only on a specific override that they will send 0-255, but it's possible there are the equivalent of the old STB DVD player "black level setting" option on many STBs, either in a menu or in a service mode. For HDMI end-devices, in fact, 16-235 appears mandatory.

Would it be reasonable to say that when an HDTV is being driven through its DVI port that by default RGB 16-235 should always be used, unless specifically overridden?

On the other hand, when a PC monitor is being driven through its DVI port that by default RGB 0-255 should always be used.

If so, would it be to much to argue that ATI's use of RGB 0-255 when their HDMI dongle is used to connect to a DVI port on an HDTV is non-standard and non-compliant with both HDTVs and regular video devices?

Essentially I'm wondering if I would be out of line sending a request to ATI's driver development team that they by default output RGB 16-235 when hooked up to an HDTV with a DVI port, just like all other regular video devices do (STBs, digital recorders, etc.).

HT Slider

06-04-08, 04:53 PM

Nothing I've seen. There might be something in the DVI 1.0 spec, but I don't have it handy.

What does Intel output when an HDTV with a DVI port is used?

RGB 0-255, suitable for a PC monitor, or RGB 16-235, suitable for an HDTV?

Thank you for your help with this whole issue.

archibael

06-04-08, 05:28 PM

Would it be reasonable to say that when an HDTV is being driven through its DVI port that by default RGB 16-235 should always be used, unless specifically overridden?

On the other hand, when a PC monitor is being driven through its DVI port that by default RGB 0-255 should always be used.

If so, would it be to much to argue that ATI's use of RGB 0-255 when their HDMI dongle is used to connect to a DVI port on an HDTV is non-standard and non-compliant with both HDTVs and regular video devices?

The specs are too vague. I found the DVI 1.0 spec and it is useless in this matter-- doesn't mention color quantization or video levels at all.

You'd have a point if it was HDMI; with DVI, I'm not seeing anything which implies 16-235 even exists.

BTW: I thought your issue was that the ATI card is sending the wrong colors levels when the dongle is not connected, and that when it is connected you're getting the 16-235 correctly.

Essentially I'm wondering if I would be out of line sending a request to ATI's driver development team that they by default output RGB 16-235 when hooked up to an HDTV with a DVI port, just like all other regular video devices do (STBs, digital recorders, etc.).

If I were you I'd eliminate the possibility of argument by asking for the option to select RGB 16-235 or RGB 0-255 on DVI. That should be doable in the registry if nothing else, but would be nice to have in the CCC.

archibael

06-04-08, 05:33 PM

What does Intel output when an HDTV with a DVI port is used?

RGB 0-255, suitable for a PC monitor, or RGB 16-235, suitable for an HDTV?

Thank you for your help with this whole issue.

Since I know of no way to differentiate between a PC monitor or a TV in the EDID, I suspect Intel sends RGB 0-255 unless told otherwise. There are some parameters in the SDK which can tweak the white and black level, but I don't see anything in the rest of the driver UI which controls this.

HT Slider

06-04-08, 05:49 PM

I looked in the registry and found the EDID for my monitor. The full extended EDID is there. I have no idea why all the tools I have tried seem not to display it.

HT I suggest you have a look in HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Enum\DISPLAY

Although the location was slightly different (probably due to running Vista), there was a complete set of EDID data in the registry.

It contains:

00 ff ff ff ff ff ff 00 52 62 00 01 62 91 02 00
1f 0d 01 03 80 00 00 00 0e ee 91 a3 54 4c 99 26
0f 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
01 01 01 01 01 01 01 1d 80 18 71 1c 16 20 58 2c
25 00 00 00 00 00 00 9e 8c 0a d0 8a 20 e0 2d 10
10 3e 96 00 00 00 00 00 00 18 00 00 00 fc 00 54
6f 73 68 69 62 61 20 54 56 0a 20 20 00 00 00 fd
00 3b 3d 1f 2e 08 00 0a 20 20 20 20 20 20 01 e2
02 02 04 04 01 1d 00 72 51 d0 1e 20 6e 28 55 00
00 00 00 00 00 1e 8c 0a a0 14 51 f0 16 00 26 7c
43 00 00 00 00 00 00 98 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 de

Running it through DTD Calculator, it comes up with:

Established Timings :
640 x 480 @ 60 Hz
Block 1 DTD Found (See List)
Block 1 DTD Found (See List)
Display does not support underscan
Display does not support basic audio
Display does not support YCbCr 4:4:4
Display does not support YCbCr 4:2:2
4 native formats in DTDs
Block 2 DTD Found (See List)
Block 2 DTD Found (See List)

Reverse calculating the DTDs in all blocks it comes up with (how did you get the full spec for each? Cut and past?):

720x480
1280x720
1920x1080

and a strange one at the very end of:

1440x480

One thing I found interesting is how DTD Calculator states that my HDTV does not support YCbCr. How does DTD Calculator figure out if YCbCr is supported or not?

Oddly, Moninfo reports slightly different information:

Monitor
Windows description......... Generic PnP Monitor
Manufacturer description.... Toshiba TV
Manufacturer................ Toshiba
————————————————————————————
Plug and Play ID............ TSB0100
Serial number............... 168290
EDID data source............ Registry (stored)
————————————————————————————
Manufacture date............ 2003, ISO week 31
EDID revision............... 1.3
Display type and signal..... Digital
Sync input support.......... n/a
Screen size................. n/a
Power management............ n/a

Color characteristics
Display gamma............... 1.00
Red chromaticity............ Rx 0.640 - Ry 0.330
Green chromaticity.......... Gx 0.300 - Gy 0.600
Blue chromaticity........... Bx 0.150 - By 0.060
White point (default)....... Wx 0.313 - Wy 0.329

Timing characteristics
VESA GTF support............ Not supported
Horizontal scan range....... 31-46kHz
Vertical scan range......... 59-61Hz
Video bandwidth............. 80MHz
Extension blocks............ 1
Timing recommendation #1.... 1920x540 at 60Hz
Modeline................ "1920x540" 74.250 1920 2008 2052 2200 540 542 547 562 +hsync +vsync
Timing recommendation #2.... 1920x1080 at 30Hz
Modeline................ "1920x1080" 74.250 1920 2008 2052 2200 1080 1084 1094 1124 interlace +hsync +vsync
Timing recommendation #3.... 720x480 at 60Hz
Modeline................ "720x480" 27.000 720 736 798 858 480 489 495 525 -hsync -vsync

Standard timings supported
640 x 480 at 60Hz - IBM VGA
720 x 480 at 60Hz - Toshiba
1920 x 540 at 60Hz - Toshiba
1920 x 1080 at 30Hz - Toshiba

Raw EDID base
00: 00 FF FF FF FF FF FF 00 52 62 00 01 62 91 02 00
10: 1F 0D 01 03 80 00 00 00 0E EE 91 A3 54 4C 99 26
20: 0F 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
30: 01 01 01 01 01 01 01 1D 80 18 71 1C 16 20 58 2C
40: 25 00 00 00 00 00 00 9E 8C 0A D0 8A 20 E0 2D 10
50: 10 3E 96 00 00 00 00 00 00 18 00 00 00 FC 00 54
60: 6F 73 68 69 62 61 20 54 56 0A 20 20 00 00 00 FD
70: 00 3B 3D 1F 2E 08 00 0A 20 20 20 20 20 20 01 E2

Raw EDID extension (reserved)
00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
40: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
50: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
60: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
70: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Display adapter
Adapter description......... ATI Radeon HD 2600 XT
Adapter device ID........... 0x95881002
Display settings............ 1360x768, 32bpp

One thing I found interesting is how DTD Calculator states that my HDTV does not support YCbCr

ToughRowToHoe

06-04-08, 05:49 PM

It seems clear that even where the extended block is not visible it must have been read, as my driver is definitely in HDMI mode (says so in CCC). Technically I guess the dongle could be forcing HDMI mode regardless of whether the HDMI vendor specific block had been read from the sink, but then I am sure ATI would have problems with using the HDMI logo and stating HDMI compliance on their products.

It is a little confusing to say the least. I am going to take a look directly in the registry later and see if I can find anything.

I don't know. I had problems getting my Onkyo receiver working with my ATI card. It would always revert to 720p on boot. They said:

"We are able to provide support only with a direction [Sic] HDMI connection to the television display. We cannot assist you getting it to work with the receiver in the middle."

As far as I know, the HDMI standard does not differentiate between HDMI interfaces. Anyhow, ATI may not care.

jong1

06-04-08, 05:52 PM

Would it be reasonable to say that when an HDTV is being driven through its DVI port that by default RGB 16-235 should always be used, unless specifically overridden?

On the other hand, when a PC monitor is being driven through its DVI port that by default RGB 0-255 should always be used.

If so, would it be to much to argue that ATI's use of RGB 0-255 when their HDMI dongle is used to connect to a DVI port on an HDTV is non-standard and non-compliant with both HDTVs and regular video devices?I think the problem is there is no way to distinguish from the EDID if a device is an HDTV DVI device or a DVI PC monitor; That is what HDMI does. Unfortunately some HDTVs came out before that was ready.

Anyway, it seems from your testing the dongle is doing what it is designed to do. At least it is forcing the driver to compress its output. As archibael says what seems more confusing in your case is why without the dongle the driver is expanding even the desktop so 0-15 on a standard bmp is clipped (then your TV clips the next 0-15, displaying only 30+). It really would be good to take a screenshot of your PC when displaying the ramp bmp and then verify with photoshop or equivalent if the bottom 0-15 has been lost due to expansion. Alternatively convert to a jpg and upload here. It would surely be odd if the desktop was still rendering the full 0-255 range of the bmp yet somehow it was being expanded and consequently clipped on-route to the TV!

jong1

06-04-08, 05:58 PM

how did you get the full spec for each? Cut and paste?I didn't bother to post my DTD timings. The things you saw there were the short video descriptions straight out of the EDID. It seems your TV does not have any of these.

archibael

06-04-08, 06:16 PM

Although the location was slightly different (probably due to running Vista), there was a complete set of EDID data in the registry.

It contains:

00 ff ff ff ff ff ff 00 52 62 00 01 62 91 02 00
1f 0d 01 03 80 00 00 00 0e ee 91 a3 54 4c 99 26
0f 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
01 01 01 01 01 01 01 1d 80 18 71 1c 16 20 58 2c
25 00 00 00 00 00 00 9e 8c 0a d0 8a 20 e0 2d 10
10 3e 96 00 00 00 00 00 00 18 00 00 00 fc 00 54
6f 73 68 69 62 61 20 54 56 0a 20 20 00 00 00 fd
00 3b 3d 1f 2e 08 00 0a 20 20 20 20 20 20 01 e2
02 02 04 04 01 1d 00 72 51 d0 1e 20 6e 28 55 00
00 00 00 00 00 1e 8c 0a a0 14 51 f0 16 00 26 7c
43 00 00 00 00 00 00 98 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 de

Running it through DTD Calculator, it comes up with:

Established Timings :
640 x 480 @ 60 Hz
Block 1 DTD Found (See List)
Block 1 DTD Found (See List)
Display does not support underscan
Display does not support basic audio
Display does not support YCbCr 4:4:4
Display does not support YCbCr 4:2:2
4 native formats in DTDs
Block 2 DTD Found (See List)
Block 2 DTD Found (See List)

Reverse calculating the DTDs in all blocks it comes up with (how did you get the full spec for each? Cut and past?):

720x480
1280x720
1920x1080

1280x720p comes from the second block, 01 1d 00 72 51 d0 1e 20 6e 28 55 00 00 00 00 00 00 1e. The other two are from the first block (what MonInfo found, below).

and a strange one at the very end of:

1440x480

That's weird, but it is indeed there: 8c 0a a0 14 51 f0 16 00 26 7c 43 00 00 00 00 00 00 98

One thing I found interesting is how DTD Calculator states that my HDTV does not support YCbCr. How does DTD Calculator figure out if YCbCr is supported or not?

The value of the fourth byte is 0x04, which translates as:

4 native formats are included as DTDs
Does not support underscan
Does not support basic audio
Does not support YCrCb 4:4:4 in addition to RGB
Does not support YCrCb 4:2:2 in addition to RGB

If you wanted to fool the system into thinking you supported YCrCb, you'd change bit 4 and/or bit 5, resulting in something like 0x14 (for 4:2:2), 0x24 (for 4:4:4) and 0x34 (for both).

Oddly, Moninfo reports slightly different information:

Timing recommendation #1.... 1920x540 at 60Hz
Modeline................ "1920x540" 74.250 1920 2008 2052 2200 540 542 547 562 +hsync +vsync
Timing recommendation #2.... 1920x1080 at 30Hz
Modeline................ "1920x1080" 74.250 1920 2008 2052 2200 1080 1084 1094 1124 interlace +hsync +vsync
Timing recommendation #3.... 720x480 at 60Hz
Modeline................ "720x480" 27.000 720 736 798 858 480 489 495 525 -hsync -vsync

Standard timings supported
640 x 480 at 60Hz - IBM VGA
720 x 480 at 60Hz - Toshiba
1920 x 540 at 60Hz - Toshiba
1920 x 1080 at 30Hz - Toshiba

MonInfo's getting all of its information from the first block, and in my opinion it is providing 1920x540 because most interlaced 1080i TVs will also do 540p. The 640x480 at gets from "standard timings" (non-DTD specified) in the first part of the block.

HT Slider

06-04-08, 06:17 PM

BTW: I thought your issue was that the ATI card is sending the wrong colors levels when the dongle is not connected, and that when it is connected you're getting the 16-235 correctly.

I have to admit I find it very difficult to explain what levels are doing what through text...

When I say I'm seeing 16-235, I mean if I take a bitmap file of a 0-255 grey level ramp and display it, that I can only see the range 16-235.

While this may seem correct, it actually is not "normal" for displaying on an HDTV with an ATI card (nor an Nvidia card for that matter). Everyone else (here at least) sees the full 0-255 when a test bitmap image of a 0-255 ramp is displayed. As a final output step, the video card is "supposed to" convert the regular 0-255 PC grey levels into YCbCr 16-235 so they look "normal" on an HDTV, just like they would on a PC monitor.

To explain further as far as video is concerned, what is "typical" is for video to be expanded from the original 16-235 into 0-255 so a PC monitor can correctly display the 16-235 grey levels contained in the video. This way when video is played back and an HDTV is hooked up, it is recompressed back (per the step above) into YCbCr 16-235 so BTB and WTW is not shown on the HDTV and instead the original 16-235 levels display as they should.

Ultimately when an HDTV is hooked up using HDMI (and the HDMI dongle), the grey levels by default look the same as if a PC monitor is used using DVI. On top of that 16-235 should be visible for video and 0-255 should be visible for photographs, desktop, games, etc.

In my case, since my HDTV doesn't actually have HDMI (but does have DVI with HDCP support), I end up in a totally different situation and need to manually fake things using both the brightness and contrast within the "Catalyst Control Center" so my output grey levels end up in line with my HDTVs 16-235 input requirement (I use +31 brightness and 73% contrast in order to get all video compliant with 16-235 visible at the same time as photographs, desktop, and video games compliant with 0-255 visible). After tons of experimenting I have been able to find a calibration where everything is calibrated correctly - but it is far from "simple" (some applications require specific brightness/contrast settings, including PowerDVD Ultra, as well as specific settings for the Catalyst Control Center).

If I simply hook up my HDTV using DVI only (with the double expansion issue) and don't make any adjustments at all within CCC, everything looks absolutely terrible. With my 0-255 test picture, I can only see roughly 30-220. Adding the HDMI dongle improves things so I see 16-235, but it is still way out of spec; until I also add the brightness/contrast adjustment. Compare this to someone hooking up an HDTV using HDMI, they simply see 0-255 when a bitmap ramp of 0-255 is displayed (of course in reality there are a number of color steps lost in the 0-255 to 16-235 compression, but these are not easy to see in a standard bitmap ramp).

I've also tried cranking up the brightness and dropping the contrast at the HDTV itself (including through the service menu) to try to get 0-255 to display. I cannot calibrate it so it displays below 16 or above 235 at all. Cranking up the brightness, 0-16 ends up a single grey and dropping the contrast enough, 235-255 ends up a single brightish grey. The grey levels seem to be digitally clipped in the internal video processor somewhere.

I hope this confusing explanation clears things up...

archibael

06-04-08, 06:20 PM

I didn't bother to post my DTD timings. The things you saw there were the short video descriptions straight out of the EDID. It seems your TV does not have any of these.

His EDID extension is version 2. SVDs weren't added until version 3.

HT Slider

06-04-08, 06:36 PM

I think the problem is there is no way to distinguish from the EDID if a device is an HDTV DVI device or a DVI PC monitor; That is what HDMI does. Unfortunately some HDTVs came out before that was ready.

ATI is already distinguishing my display as an HDTV and not a PC monitor. Catalyst Control Center, by default, detects it as a Toshiba TV with HDTV support of 720p and 1080i. It also already goes into the HDTV mode where it automatically displays all regular PC resolutions into standard HDTV formats (720p and 1080i). The "treat as HDTV" box isn't even there, since it knows it is an HDTV.

When I hook up a PC monitor and reboot I see a completely different set of options, no HDTV support (unless I click the "treat as HDTV" check box), and the HDTV mode where resolutions are scaled into supported ones is turned off (only resolutions supported by the monitor are available).

The video card must be taking the fact that no standard PC formats are supported, yet all standard HDTV formats (up to 1080i) are supported as justification to turn on "HDTV display mode". Unfortunately, while it is smart enough to do that, it isn't smart enough to realize that HDTVs require grey levels between 16-235, regardless if DVI or HDMI is used.

EDIT: Also, the display manager also shows a picture of a TV and calls it a TV vs when I plug in my Viewsonic monitor, instead it shows a picture of a monitor and calls it a monitor.

Grey levels seem to be something ATI simply overlooked when recognizing a display as a TV, but it has a DVI port on it. They did at least get most of it right (turning on HDTV mode automatically).

Anyway, it seems from your testing the dongle is doing what it is designed to do. At least it is forcing the driver to compress its output. As archibael says what seems more confusing in your case is why without the dongle the driver is expanding even the desktop so 0-15 on a standard bmp is clipped (then your TV clips the next 0-15, displaying only 30+). It really would be good to take a screenshot of your PC when displaying the ramp bmp and then verify with photoshop or equivalent if the bottom 0-15 has been lost due to expansion. Alternatively convert to a jpg and upload here. It would surely be odd if the desktop was still rendering the full 0-255 range of the bmp yet somehow it was being expanded and consequently clipped on-route to the TV!

I have taken screenshots and they always look 100% identical to the original bitmap image that I am displaying. All of this expansion and compression is going on at what appears to be a final step, right before being output from the video card. It is as if the video card is able to do expansion and compression just before output and in my case I can get a false extra expansion if DVI to DVI is used (only with ATI cards though; Nvidia cards behave the same as if I use the HDMI dongle on the ATI card).

The good news is Catalyst Control Center's "color" settings also only affect the final output too. Here I am able to compensate and get the correct output levels and it doesn't affect the numerical values contained within a screenshot in any way. The only settings that do affect screenshot levels are the AVIVO brightness/contrast settings and only when video is played. I would assume PowerDVD's own brightness/contrast settings also affect screenshot levels.

archibael

06-04-08, 06:42 PM

Okay, HTSlider, I think I understand where you're coming from, now: your video card is not re-compressing to 16-235-- which is what your TV expects-- and for that matter it's not leaving it de-compressed at 0-255.

Perhaps your TV clips any 0-255 input it receives over DVI to 16-235... and ATI is clipping to 16-235 already, resulting in double clipping; whereas with the dongle ATI is not clipping and the TV is the only problematic device.

I have no idea. I'm honestly making crap up at this point :o since I don't know enough about the ATI interaction with your monitor.

HT Slider

06-04-08, 07:05 PM

If you wanted to fool the system into thinking you supported YCrCb, you'd change bit 4 and/or bit 5, resulting in something like 0x14 (for 4:2:2), 0x24 (for 4:4:4) and 0x34 (for both).

With Vista, do you think that simply editing the registry EDID data would change how the ATI driver treats the display?

I guess faking support for YCbCr alone still wouldn't actually turn on YCbCr. I'd probably have to populate the E-DID with HDMI compliant data as well.

Even then, if I reboot, would the registry be automatically overwritten by a fresh reading of EDID from the display again?

I seem to recall there being some custom service that could be installed that would populate the EDID in the registry and keep it populated that way to work around similar issues with Nvidia cards (5+ years ago). This was one of the earlier ideas for getting around EDID issues; before video drivers had the ability to override EDID and add HD formats (and without using Powerstrip).

I wonder if Powerstrip could somehow toggle between YCbCr vs RGB output modes?

HT Slider

06-04-08, 07:21 PM

Okay, HTSlider, I think I understand where you're coming from, now: your video card is not re-compressing to 16-235-- which is what your TV expects-- and for that matter it's not leaving it de-compressed at 0-255.

Correct; although it is simply leaving it decompressed at 0-255 when I use the HDMI dongle so for me the second expansion is no longer an issue (although it does appear to be an ATI bug that probably affects others).

Perhaps your TV clips any 0-255 input it receives over DVI to 16-235... and ATI is clipping to 16-235 already, resulting in double clipping; whereas with the dongle ATI is not clipping and the TV is the only problematic device.

I have no idea. I'm honestly making crap up at this point :o since I don't know enough about the ATI interaction with your monitor.

The TV definitely behaves as if it clips 0-255 to 16-235 and that seems like the most logical explanation as far as the HDTV is concerned.

I don't think ATI is clipping it, but instead is getting confused and expanding instead of compressing the grey levels for some reason when DVI to DVI is used. Perhaps since they know it is an HDTV hooked up using DVI, they know they need to either expand or compress and unfortunately they got it backwards and did the wrong thing. They probably don't have an HDTV in house with a DVI input so they wouldn't have tested this scenario.

The more I think about it, this is the one and only situation where RGB is used to feed an HDTV (ATI would expect me to use DVI to hook up my HDTV). Very likely they did simply get the math backwards and perform an expansion instead of a compression. In this one and only case they need to compress RGB 0-255 into RGB 16-235 for output. With HDTVs hooked up using HDMI, they would be less likely to get it backwards since the transform is from RGB 0-255 to YCbCr 16-235 (for HD resolutions anyway - BT.709).

The guys doing the coding probably thought they were doing the right thing by doing an RGB to RGB grey level conversion and threw it in; only to make an error and get it backwards.

Luckily when I use the HDMI dongle, it seems to disable this extra expansion so I only have 1/2 of the problem.

Wo0zy

06-04-08, 07:29 PM

With Vista, do you think that simply editing the registry EDID data would change how the ATI driver treats the display?

I guess faking support for YCbCr alone still wouldn't actually turn on YCbCr. I'd probably have to populate the E-DID with HDMI compliant data as well.

Even then, if I reboot, would the registry be automatically overwritten by a fresh reading of EDID from the display again?

I seem to recall there being some custom service that could be installed that would populate the EDID in the registry and keep it populated that way to work around similar issues with Nvidia cards (5+ years ago). This was one of the earlier ideas for getting around EDID issues; before video drivers had the ability to override EDID and add HD formats (and without using Powerstrip).

I wonder if Powerstrip could somehow toggle between YCbCr vs RGB output modes?

From experimentation we did for other reasons, it seems that any manual changes to the registry entries are overwritten with the "real" EDID from the display on every reboot (re-detection) as you suggest.

It might be worth researching a custom service or perhaps looking into updating the displays EDID (assuming the EPROM isn't write protected).

I don't think Powerstrip can switch output modes but it might be worth experimenting with the "color profiles" configuration options.

Wo0zy

HT Slider

06-04-08, 09:06 PM

From experimentation we did for other reasons, it seems that any manual changes to the registry entries are overwritten with the "real" EDID from the display on every reboot (re-detection) as you suggest.

It might be worth researching a custom service or perhaps looking into updating the displays EDID (assuming the EPROM isn't write protected).

I don't think Powerstrip can switch output modes but it might be worth experimenting with the "color profiles" configuration options.

Wo0zy

My previous experiments doing this also showed the registry entries being overwritten on every reboot.

It was a while ago, but I do recall both figuring out a temporary method for experiments as well as other people successfully writing new EDID data to Toshiba HDTVs of the same series as mine.

archibael

06-05-08, 01:26 AM

The pay version of Powerstrip can rewrite the EPROM which holds the EDID on some TVs.

djos

06-05-08, 02:07 AM

The pay version of Powerstrip can rewrite the EPROM which holds the EDID on some TVs.

Feel free to elaborate on how to do this, I've just acquired v3.76 and am using it to force 1080p@50 on my HD 2600 PRO.

archibael

06-05-08, 04:16 AM

Feel free to elaborate on how to do this, I've just acquired v3.76 and am using it to force 1080p@50 on my HD 2600 PRO.

Evidently there is an Update EDID button on the registered version. Here is an example for how it was used, though the method for un-write protecting the EEPROM probably differs from manufacturer to manufacturer.

http://forums.entechtaiwan.net/viewtopic.php?t=5984

jong1

06-05-08, 05:20 AM

ATI is already distinguishing my display as an HDTV and not a PC monitor. Catalyst Control Center, by default, detects it as a Toshiba TV with HDTV support of 720p and 1080i. It also already goes into the HDTV mode where it automatically displays all regular PC resolutions into standard HDTV formats (720p and 1080i). The "treat as HDTV" box isn't even there, since it knows it is an HDTV.OK. Fair comment, but I think this just means HDTV in ATI speak means a device that can do 720p/1080i/1080p.

What I mean is there is no way for it to know this is one of the halfway house DVI enabled TVs (where 16-235 might be more appropriate) rather than a DVI monitor supporting those resolutions (where someone might want 0-255).

Anyway I certainly agree with the idea that we should be able to select this in CCC, rather than have it forced on us by some combination of dongle, device type and phases of the moon!

I have taken screenshots and they always look 100% identical to the original bitmap image that I am displaying. All of this expansion and compression is going on at what appears to be a final step, right before being output from the video card.Wow that is odd. As you say it looks like ATI may just have got their maths backwards! Definitely worth raising a bug report on that. I would ignore all the dongle stuff, which seems to be working as it should and just say how you are getting an expanded desktop when displaying bitmaps on your DVI TV, using default settings and no dongle.

Maybe now though you can agree you are indeed seeing "strange" behaviour (http://www.avsforum.com/avs-vb/showthread.php?p=13963667#post13963667) ;):)

Wo0zy

06-05-08, 05:47 AM

So what do we think would happen if you were to connect a PC monitor with an HDMI port to an ATI card? Where's quite a few about these days.

If a typical user received the ATi dongle with the PC/ Graphics Card, logic would suggest that they would use it to connect to the monitor via HDMI. I guess with an RGB capable display it could be calibrated correctly for video levels without "loss" if the dongle forced compression but is that really the point?

If anyone following this thread has one and would care to comment and maybe post the EDID, it would be very interesting.

Wo0zy

jong1

06-05-08, 09:16 AM

That is an interesting point. If you read that lovely HDMI spec all output to an HDMI device should be "limited range" (16-235), except for old-fashioned standard VGA, even when outputting RGB.

Now you could argue this matters little if you calibrate for limited range, but I am sure those doing professional artwork/photo editing do not want their high quality images compressed in range just to meet the HDMI spec.

I bet most display buyers do not appreciate these subtleties. More likely their monitor has some built in speakers and HDMI offers "one cable simplicity". We can only hope those who are buying for professional or even semi-pro use are aware, but in many cases I doubt it!

HT Slider

06-05-08, 02:41 PM

What I mean is there is no way for it to know this is one of the halfway house DVI enabled TVs (where 16-235 might be more appropriate) rather than a DVI monitor supporting those resolutions (where someone might want 0-255).

Maybe now though you can agree you are indeed seeing "strange" behaviour (http://www.avsforum.com/avs-vb/showthread.php?p=13963667#post13963667) ;):)

I don't think there is much point in insulting the HDTV. The problem isn't the HDTV. It is that HDTVs are not being driven correctly by todays video cards and Media Center PCs.

Back in 2002/2003/2004 (which isn't really all that long ago for a TV since TV's are typically used for a good 10 years), all high end HDTVs had DVI ports on them. Nothing had HDMI (was the spec even written ?). These high end HDTVs typically sold for $4000-$5000 at the time and still produce a very good 1920x1080 resolution image today (when calibrated inputs are used).

The fact that any (low to high end) hardware video device (STB, etc.) can communicate with these more than 3 year old HDTVs properly says to me that it would be reasonable to expect that a "high end" Media Center PC should also be able to communicate properly with these HDTVs.

Keep in mind that Media Center PCs are not being marketed as do it yourself, home built HTPCs and as such should work as well as a cheap STB when an HDTV is used, regardless if the TV is more than 3 years old.

On top of that, it would be relatively simple for drivers to be updated to get this right so "regular" customers can simply plug and play.

Major issues today:

- SD video sources (BT.601) are not expanded by default with ATI drivers, yet Nvidia drivers do expand. With ATI we have to add "UseBT601CSC=1" in the registry to get SD to expand just like HD does by default.
- If an HDTV with an HDMI input is hooked up using a totally spec compliant DVI to HDMI cable and the Media Center has an ATI card, the HDTV is fed RGB 0-255, even though the EDID contains the correct information, per the HDMI spec, to tell the video card to send YCbCr 16-235. Since HDTVs, when fed RGB, require 16-235, grey levels are totally messed up.
- If an HDTV with a DVI input is used using a direct DVI cable, ATI drivers erroneously expand the grey levels on output, instead of compressing (should be 0-255 mapped into 16-235, but instead they do the opposite). Nvidia drivers on the other hand simply output the original RGB 0-255 and this is also wrong since all HDTVs are both detected by the video card/driver and HDTVs all require 16-235.
- Nvidia doesn't perform grey level expansion with all renderers so grey levels vary depending on the application being run.

Between these, there is not a single Media Center PC out there anywhere that by default provides correct grey levels with all software when hooked up to an HDTV today. Nvidia cards do actually provide correct grey levels when Media Center itself and PowerDVD Ultra are used, but Media Player and other software doesn't provide correct grey levels. With ATI cards we need to resort to registry hacks and depending on the way it is hooked up, manually adjusting the brightness and contrast in CCC. At least with ATI cards we can ultimately "adjust" things into compliance, but this is ridiculous.

Basically these Media Center PCs are a total mess when it comes to driving HDTVs and this is exactly what they are being marketed and sold for. It is totally reasonable not to be pleased with this and also totally reasonable to expect ALL HDTVs to be driven using the correct grey levels, including ones that are older than 3 years old.

Archibael, how is Intel handling this stuff? Has Intel made a concerted effort to ensure their Media Center targeted hardware will drive HDTVs correctly?

We really need one of the video device makers to get a handle on this so turn key, functional, Media Centers can be made available. Market pressure, at that point, would probably encourage the rest to eventually get things right too.

jong1

06-05-08, 03:02 PM

I don't think there is much point in insulting the HDTV. The problem isn't the HDTV. It is that HDTVs are not being driven correctly by today's video cards and Media Center PCs.I wasn't insulting the TV, just teasingly reminding you that we all now think it is the combination of your TV and the card that is leading to your problem rather than a Vista driver bug!

Anyway, I agree there is a lot of room for improvement in driver quality and consistency when it comes to HDTVs. Also, they need to provide much more control (at least for advanced users) instead of making arbritrary undocumented decisions, which are certain to be wrong for some users.

arfster

06-05-08, 03:04 PM

We really need one of the video device makers to get a handle on this so turn key, functional, Media Centers can be made available. Market pressure, at that point, would probably encourage the rest to eventually get things right too.

They don't care, to be blunt. Until one of the major review sites slags them off for getting this wrong, they'll continue to not care.

Unfortunately, major review sites are usually pretty clueless about this too, and typically don't even notice the problem.

archibael

06-05-08, 03:20 PM

Archibael, how is Intel handling this stuff? Has Intel made a concerted effort to ensure their Media Center targeted hardware will drive HDTVs correctly?

We really need one of the video device makers to get a handle on this so turn key, functional, Media Centers can be made available. Market pressure, at that point, would probably encourage the rest to eventually get things right too.

Unfortunately, I'm pretty much a hack. My IGP knowledge is all self-taught-- I'm not involved with the graphics driver teams at all, so I can't say what the overall strategy has been.

I will say that, in general, Intel software engineers are sticklers for The Spec and The License Agreement above all else-- and since this sometimes trumps ease of customer use, this can be annoying (as in the case of HDCP-repeater mode and Denon receivers prior to to the 3808). However, they do tend to immediately fix something which is not done by spec if you can show them that they are in violation.

The trick in getting Intel graphics guys to fix their drivers lies in figuring out what is going on, and proving it's really a driver/HW bug and not software.

And arfster is very spot on in his comments, I've gotta say.

crabnebula

06-05-08, 09:48 PM

Major issues today:

- SD video sources (BT.601) are not expanded by default with ATI drivers, yet Nvidia drivers do expand. With ATI we have to add "UseBT601CSC=1" in the registry to get SD to expand just like HD does by default.
- If an HDTV with an HDMI input is hooked up using a totally spec compliant DVI to HDMI cable and the Media Center has an ATI card, the HDTV is fed RGB 0-255, even though the EDID contains the correct information, per the HDMI spec, to tell the video card to send YCbCr 16-235. Since HDTVs, when fed RGB, require 16-235, grey levels are totally messed up.
- If an HDTV with a DVI input is used using a direct DVI cable, ATI drivers erroneously expand the grey levels on output, instead of compressing (should be 0-255 mapped into 16-235, but instead they do the opposite). Nvidia drivers on the other hand simply output the original RGB 0-255 and this is also wrong since all HDTVs are both detected by the video card/driver and HDTVs all require 16-235.
- Nvidia doesn't perform grey level expansion with all renderers so grey levels vary depending on the application being run.

Between these, there is not a single Media Center PC out there anywhere that by default provides correct grey levels with all software when hooked up to an HDTV today. Nvidia cards do actually provide correct grey levels when Media Center itself and PowerDVD Ultra are used, but Media Player and other software doesn't provide correct grey levels. With ATI cards we need to resort to registry hacks and depending on the way it is hooked up, manually adjusting the brightness and contrast in CCC. At least with ATI cards we can ultimately "adjust" things into compliance, but this is ridiculous.

Basically these Media Center PCs are a total mess when it comes to driving HDTVs and this is exactly what they are being marketed and sold for. It is totally reasonable not to be pleased with this and also totally reasonable to expect ALL HDTVs to be driven using the correct grey levels, including ones that are older than 3 years old.

Good summary of issues. It should be pointed out however that some HDTVs (such as my Samsung) have an option for selecting 16-235 or 0-255 RGB with HDMI. And if you also use your HTPC as a PC, you might prefer having a single 0-255 calibration on the set for viewing the desktop, which means you'll want your video expanded.

In the end, the solution to this mess is so simple that I don't understand why nVidia and ATI don't implement it. Just let us choose manually what kind of output we want (like the PS3 apparently does) and compress the desktop or expand the video accordingly.

Anyway I certainly agree with the idea that we should be able to select this in CCC, rather than have it forced on us by some combination of dongle, device type and phases of the moon!

tdanco

06-06-08, 01:18 AM

My ati flag is not waving!?

I also cannot watch dvds properly without an annoying horizontal tear across the screen on faster moving scenes. It is very frustrating. BD's work fine. I have an msi 2600xt and a sharp aquos lcd42d64u tv.

Any input would be appreciated.

HT Slider

06-06-08, 01:45 AM

Good summary of issues. It should be pointed out however that some HDTVs (such as my Samsung) have an option for selecting 16-235 or 0-255 RGB with HDMI. And if you also use your HTPC as a PC, you might prefer having a single 0-255 calibration on the set for viewing the desktop, which means you'll want your video expanded.

In the end, the solution to this mess is so simple that I don't understand why nVidia and ATI don't implement it. Just let us choose manually what kind of output we want (like the PS3 apparently does) and compress the desktop or expand the video accordingly.

I totally agree with you, however the default when outputting to an HDTV should be 16-235 and there should be a toggle in the driver control panel that lets us override the default and switch between RGB 0-255, RGB 16-235, or YCbCr.

To send an HDTV 0-255 by default is just wrong no matter how it is hooked up.

I do agree, in the rare case that a TV supports it, that RGB 0-255 might be preferred in some situations (not very many though).

jong1

06-06-08, 04:06 AM

Wo0zy

06-06-08, 06:48 AM

What's even stranger is that your Moninfo report (and mine IIRC) claims the EDID source is "Registry (Stored)" but it can't be as your registry contains the extension block. I also can't figure out why the RivaTuner report gave me a complete EDID but didn't work for you. Did you get are chance to hook up the HTPC directly to the TV to see if the AVR is causing a problem?

HT, I'm borrowing one of our older test displays this weekend (it's a 3-4 year old 32" LG LCD HDTV with DVI). Obvioulsy there's no guarantee it will behave the same as your setup but I thought it would be interesting to see.

I'll do a few tests of my own then download your test material as well. I think you posted it a few pages back but if you've got anything else you'd like checked, let me know.

Wo0zy

jong1

06-06-08, 07:17 AM

What's even stranger is that your Moninfo report (and mine IIRC) claims the EDID source is "Registry (Stored)" but it can't be as your registry contains the extension block. I also can't figure out why the RivaTuner report gave me a complete EDID but didn't work for you. Did you get are chance to hook up the HTPC directly to the TV to see if the AVR is causing a problem?No I haven't. Might try it over the weekend. Might need a stiff drink first though. I never like to touch anything when all is working well. I'm not convinced it is the AVR now though, because others without AVRs seem to have similar problems. There were a lot of entries in the registry. It was quite hard to find the right one. I will try to scan through them all to see if one is incomplete. It may be the Moninfo is simply pulling the wrong entry.

jong1

06-06-08, 10:12 AM

It all gets even more confusing.

It turns out my registry had about a dozen different keys all under the same "master" key for my Denon amp. Some had the full EDID, some the truncated version. It happens that the key that windows claimed to be using (the one with a "control" key and the one referred to if I look at the monitor in device manager) had a truncated EDID. It had a device description of "plug and play monitor" instead of the device specific description of "DENON-AVAMP", which is contained in the EDID and which was the Device Description in the key that had the full EDID.

There were also other keys for various screens I have connected over the last 18 months since i first installed Windows.

I uninstalled the monitor in device manager, deleted all the sub-keys under Enum\DISPLAY and rebooted. I then had only one key under the "DON004" key for my monitor and it has truncated info!

I tried rebooting, turning the PC on without the AV amp and then turning on the amp, turning on the amp with my Cable STB selected waiting several minutes and then selecting the PC input, but still I have only one key for my Denon amp and it is truncated. I have no idea how I had built up about a dozen in the past and why some were full and some truncated. It occurred to me later that this could well be because previously I had an Nvidia 7600GT card. Maybe this gathered the full EDID data and now a different registry key is being used.

I then connected the TV directly and the new EDID is also truncated. I tried to read this EDID using Moninfo and Rivatuner. Moninfo showed the old Denon info, even when I "refreshed" it. Rivatuner just said it was "Unable to read EDID". Admittedly I had not rebooted after switching the HDMI over.

All is working fine, which seems unlikely if the extended data was not being read at some level, but I don't understand how so many multiple keys for the same device had built up or why the EDID is now truncated. I guess it is possible that a new one appears each time a new driver is installed and the full EDID's come from when my Nvidia card was installed. But still the ATI card must be reading the full data or I am pretty sure things would not be working properly.

:confused::confused:

Wo0zy

06-06-08, 10:55 AM

archibael

06-06-08, 11:47 AM

@archibael

I just read your post in another thread that said Moninfo directly interrogates the sink device to gather EDID data.

This being the case do you have anty explanation, or even idea, why Moninfo seems to return a blank second block, when the full extended data is clearly in the registry?

MonInfo sends a query to the graphics driver, which interrogates the sink device. So it's either a bug in MonInfo's call to the driver, or in the driver's report-out to the query.

jong1

06-06-08, 12:40 PM

This is strange behaviour indeed.

The only things that spring to mind are the damn dongle and possibly differences in ATI drivers. Pure speculation and totally unfounded. More research me thinks.

Cheers,

Wo0zyI have substantially edited my post above. I thought there was an extended block for my TV when directly connected but I was mistaken. Also, it occurred to me (should have happened earlier :o) that the full keys may have come from when I had an Nvidia card installed!

jong1

06-06-08, 12:56 PM

Hey, just done a trawl through my registry looking for anything EDID related and I found this in HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Video\{E C5703C4-95A2-4EB3-A777-565D31CB8D61} (my 3850 registry entries)

"DALRULE_GetTVFakeEDID" = "1"

Interesting? Anyone up for turning it off?!

HT Slider

06-06-08, 01:26 PM

HT, I'm borrowing one of our older test displays this weekend (it's a 3-4 year old 32" LG LCD HDTV with DVI). Obvioulsy there's no guarantee it will behave the same as your setup but I thought it would be interesting to see.

I'll do a few tests of my own then download your test material as well. I think you posted it a few pages back but if you've got anything else you'd like checked, let me know

Cool. I am very interested to hear your results.

I have been sortof holding off on pushing this too hard with both ATI and Microsoft because I've had this fear that perhaps my HDTV somehow isn't a "normal" DVI HDTV.

Being LG (very much a PC device manufacturer), do you think it is more likely than my Toshiba for supporting RGB 0-255?

I guess we'll find out if it supports RGB 0-255 or RGB 16-235 in a couple of days.

(Edited a little) The only test you really need to do is to simply display the 0-255 bitmap ramp using the HDMI dongle and default settings and see if you can see the whole thing using default brightness/contrast settings everywhere (with a normally calibrated TV; calibrated for RGB 16-235). I can only see 16-235 when using default settings. When I then plug in my LG or Viewsonic PC monitor instead I can see 0-255 without changing any settings anywhere (EDIT: I actually did this test with the Nvidia card installed, but I suspect since my HDTV was displaying the same 16-235 prior to hooking up the monitors that it should be the same).

Assuming it is the same, it would be interesting to know if by adjusting the brightness/contrast on the HDTV if it is possible to calibrate it for RGB 0-255. I know mine seems to digitally clip RGB at 16-235 so I can not ever display WTW or BTB through a DVI source. I find this odd and would expect the full 0-255 to be maintained so I should be able to at least get a little WTW and BTB to display (but I can't).

If you like, you could also confirm that when UseBT601CSC=1 is set and the CCC "color" form's brightness and contrast are set to +31/73% respectively that everything from pictures (0-255 visible), SD video (16-~235+ visible) and HD video (16-235 visible) is all within spec. when the display is calibrated for 16-235 RGB. This is how I currently have everything set. I also use +19/-5 brightness/contrast within PowerDVD in order to get the Blu-ray and HD-DVD calibration disks to display to spec (16-235 visible).

EDIT: After that, it would be interesting to see if when using a direct DVI to DVI cable and default settings if you too can only see roughly 30-220 of the same 0-255 ramp with the display calibrated for RGB 16-235.

HT Slider

06-06-08, 01:28 PM

I have substantially edited my post above. I thought there was an extended block for my TV when directly connected but I was mistaken. Also, it occurred to me (should have happened earlier :o) that the full keys may have come from when I had an Nvidia card installed!

Interesting idea. My full EDID may also have been when I had an Nvidia card installed I suppose too.

Wo0zy

06-06-08, 01:58 PM

Interesting idea. My full EDID may also have been when I had an Nvidia card installed I suppose too.

Umm. Mine too :o

I've also brought home a new system which is ATI based (admittedly IGP) which has only ever been connected to a VGA monitor in the office (for PDI). I'll hook that up to my TV (and the older LG) a bit later and see what EDID results I get with that using Moninfo, EDID Viewer and RivaTuner.

Also, without a bit of testing I have no Idea what formats the old LG supports but I suspect you're right. Will keep you posted. I intend to use the "Ramp" video that Arfster recommends and will also use your 0-255 BMP for the primary tests.

I noticed earlier that there are a couple of other active threads here regarding ATI graphics/chipsets and possible EDID issues (although they're not identified as such yet). I wonder if we're on to something.

@jong1,

I'll have a look for that registry key as well. It's interestingly named for something that's not relevant don't you think?

Cheers,

Wo0zy

HT Slider

06-06-08, 05:50 PM

Umm. Mine too :o

I've also brought home a new system which is ATI based (admittedly IGP) which has only ever been connected to a VGA monitor in the office (for PDI). I'll hook that up to my TV (and the older LG) a bit later and see what EDID results I get with that using Moninfo, EDID Viewer and RivaTuner.

Also, without a bit of testing I have no Idea what formats the old LG supports but I suspect you're right. Will keep you posted. I intend to use the "Ramp" video that Arfster recommends and will also use your 0-255 BMP for the primary tests.

I noticed earlier that there are a couple of other active threads here regarding ATI graphics/chipsets and possible EDID issues (although they're not identified as such yet). I wonder if we're on to something.

@jong1,

I'll have a look for that registry key as well. It's interestingly named for something that's not relevant don't you think?

Cheers,

Wo0zy

I also used the ramp Arfster provided the link to. I also created a Media Center dvr-ms recording from it so I could test the calibration playing directly from Recorded TV in Media Center.

Once I used the HDMI adapter, UseBT601CSC=1 and cranked the "color" brightness/contrast to +31/73%, everything to do with Media Center and Media Player in SD and HD, including Arfster's recommended ramp and the dvr-ms ramp were all properly displaying 16-235 (and at the same time 0-255 for bitmaps). The only thing this didn't fix was PowerDVD, where I needed to calibrate it using separate Bluray/HD-DVD calibration DVDs (and ended up with the +19/-5 brightness/contrast).

BTW, the bmp and wmv ramps I attached in that earlier post were not created by me; they were created by Ian Kennedy a member of Microsoft's eHome (Media Center) software development team (TV playback quality group).

crabnebula

06-06-08, 10:52 PM

I totally agree with you, however the default when outputting to an HDTV should be 16-235 and there should be a toggle in the driver control panel that lets us override the default and switch between RGB 0-255, RGB 16-235, or YCbCr.

To send an HDTV 0-255 by default is just wrong no matter how it is hooked up.

I do agree, in the rare case that a TV supports it, that RGB 0-255 might be preferred in some situations (not very many though).

So say we all!

Beamer

06-07-08, 07:22 AM

djos

06-07-08, 07:50 AM

Some of you folk need to read this Wiki on YCbCr which is the standard Colour Space for Digital Video (16-235): http://en.wikipedia.org/wiki/YCbCr

It'll prolly make a few things clearer.

:)

warlockuk

06-07-08, 09:56 AM

I've got a GeCube 2400Pro running whatever the latest driver is (8.453A or something - I think it's based on the March ATI driver) and it took months to actually get the machine to accelerate HD-MKVs in hardware...

...and now the problem I'm having is that while viewing hi-def video it'll lockup and crash (and then restart) after a few minutes of playback.

When I install that ATI tool and go to the overclocking options to do a GPU test (the thing with the rotating space ship and an artifact checker) it runs fine with the artifact checker - until you click the GPU button to test that. Then after about 4 seconds testing the screen goes blank and the machine crashes...

The fan is working fine on the card (though the card doesn't seem to report speed or temperatures unfortunately)...

...has anyone else had a similar experience? I'm wondering if it is the driver or if the card is just unsuitable for use as a graphics card. In which case I'm going to contact eBuyer and ask to upgrade it to a 3450 or something...

Cheers.

arfster

06-07-08, 10:11 AM

Wo0zy

06-07-08, 01:48 PM

HT,

I don't think any testing I do is going to help by adding further evidence to your case unfortunately,

The EDID for the LG is as follows

00 ff ff ff ff ff ff 00 1e 6d 46 75 01 01 01 01
02 10 01 03 81 40 26 78 08 b1 da a1 56 48 98 24
13 48 4b a1 08 00 31 40 01 01 01 01 45 40 01 01
61 40 81 80 01 01 4e 1f 00 90 51 00 1b 30 40 88
13 00 a2 0b 32 00 00 18 1b 21 50 a0 51 00 1e 30
48 88 35 00 a2 0b 32 00 00 1c 00 00 00 fd 00 3b
3d 1f 30 09 00 0a 20 20 20 20 20 20 00 00 00 fc
00 52 5a 33 32 4c 5a 35 35 0a 20 20 20 20 00 56
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

You'll notice that byte 126 is "00" indicating that there is no extension block. As a result CCC sees the display as a "Digital Panel" rather than a "DTV".

This is the case regardless of whether the ATI dongle is used (via an HDMI>DVI cable) or not (straight DVI) and as a result both connections types need calibrating for the same levels (0-255).

Sorry :(.

Wo0zy

HT Slider

06-07-08, 03:59 PM

HT,

I don't think any testing I do is going to help by adding further evidence to your case unfortunately,

...

You'll notice that byte 126 is "00" indicating that there is no extension block. As a result CCC sees the display as a "Digital Panel" rather than a "DTV".

This is the case regardless of whether the ATI dongle is used (via an HDMI>DVI cable) or not (straight DVI) and as a result both connections types need calibrating for the same levels (0-255).

Sorry :(.

Wo0zy

That is to bad. I was hoping for confirmation. Note that I noticed mine is now also listed as "Digital Panel" with 8.5 and the HDMI dongle in use, but for certain "DTV" with previous drivers and DVI to DVI. I can't say for absolute certain, but I'm fairly sure 8.5 also labels it "DTV" when DVI to DVI is used.

Still, I am interested in some of the performance aspects. As long as the ATI card still turns on HDTV mode I still feel it should compress the output into 16-235.

Does it turn on HDTV mode? Or does it only enable 640x480, 1280x720, or 1920x1080 (and/or garble the display with other PC resolutions) and treat it like a monitor?

In all situations where the 2600XT digitally drives my Toshiba 51H83 (regardless of what it labels it as), the video card automatically enables both 720p and 1080i, automatically creates custom resolutions to compensate for overscan, automatically enables and scales all PC resolutions into one of the custom, overscan compensated resolutions, and automatically outputs all resolutions using either 720p or 1080i timings. This is what I consider "HDTV mode".

Without "HDTV" mode being enabled, overscan isn't automatically compensated for and only 640x480, 1280x720 and 1920x1080 actually display anything. I can create this situation by using my YPbPr transcoder (the ATI card assumes it is a monitor). I also need to manually enable "Treat as HDTV" and enable 720p + 1080i, but it still doesn't turn on HDTV mode (no overscan compensation nor rescaling of standard PC resolutions into HDTV resolutions). Note though that I haven't actually used the YPbPr transcoder since somewhere around Catalyst 7.6 (so I don't know if the latest versions truly enable HDTV mode by checking "Treat as HDTV").

If the LG HDTV is recognized as a monitor, HDTV mode shouldn't be enabled.

Sorry I can't be more accurate on the labels... I was always more focused on performance rather than what ATI labelled it as. I guess I should pull out the HDTV and the HTPC again and thread through the old DVI cable through to confirm what 8.5 labels it as. I'm fairly sure it is DTV, just like it is with 8.3.

Wo0zy

06-07-08, 06:32 PM

That is to bad. I was hoping for confirmation. Note that I noticed mine is now also listed as "Digital Panel" with 8.5 and the HDMI dongle in use, but for certain "DTV" with previous drivers and DVI to DVI. I can't say for absolute certain, but I'm fairly sure 8.5 also labels it "DTV" when DVI to DVI is used.

Still, I am interested in some of the performance aspects. As long as the ATI card still turns on HDTV mode I still feel it should compress the output into 16-235.

Does it turn on HDTV mode? Or does it only enable 640x480, 1280x720, or 1920x1080 (and/or garble the display with other PC resolutions) and treat it like a monitor?

In all situations where the 2600XT digitally drives my Toshiba 51H83 (regardless of what it labels it as), the video card automatically enables both 720p and 1080i, automatically creates custom resolutions to compensate for overscan, automatically enables and scales all PC resolutions into one of the custom, overscan compensated resolutions, and automatically outputs all resolutions using either 720p or 1080i timings. This is what I consider "HDTV mode".

Without "HDTV" mode being enabled, overscan isn't automatically compensated for and only 640x480, 1280x720 and 1920x1080 actually display anything. I can create this situation by using my YPbPr transcoder (the ATI card assumes it is a monitor). I also need to manually enable "Treat as HDTV" and enable 720p + 1080i, but it still doesn't turn on HDTV mode (no overscan compensation nor rescaling of standard PC resolutions into HDTV resolutions). Note though that I haven't actually used the YPbPr transcoder since somewhere around Catalyst 7.6 (so I don't know if the latest versions truly enable HDTV mode by checking "Treat as HDTV").

If the LG HDTV is recognized as a monitor, HDTV mode shouldn't be enabled.

Sorry I can't be more accurate on the labels... I was always more focused on performance rather than what ATI labelled it as. I guess I should pull out the HDTV and the HTPC again and thread through the old DVI cable through to confirm what 8.5 labels it as. I'm fairly sure it is DTV, just like it is with 8.3.

Not sure where to start.

Firstly this system is using 8.4 (will happily update it to 8.5 but I doubt it'll effect things). Under "Display Manager" by default I'm only offered the resolutions available in the EDID. If I go to the "Digital Panel" options and look under "HDTV Support", the bottom selection box offers no HDTV timings at all. If I tick an HDTV timing in the top box (to force the timing to be available in display manager) it does become selectable both in display manager and in the lower box in "HDTV Support". This TV doesn't like 1080i (shakes like hell) but 720p works fine. With 720p "forced" I can use either the full 1280x720 or "optimised" 1152x648 timing or even "add" a new timing to compensate for overscan (resulting in 1168x676 active pixels within a standard 720p timing). Regardless, selecting any of these timings has no impact on the color formats or greyscale levels with either the dongle in-line or using a straight DVI>DVI connection.

I'm not really sure about your definition of HDTV mode enabled? Like I said, by default I'm offered the resolutions found in the EDID (which are PC resolutions). If I then force an HDTV timing only the HDTV timings are offered in display manager until I again "force" a PC resolution then I'm once again offered the EDID timings. I guess this means it's switching between the two modes?

TBH I'm not really sure where this leaves things. I'm sure that, by default, this display is treated as a monitor but I can force HDTV timings if desired but neither scenario changes the output as far as I can see (except the resolution and therefore scaling).

It's likely that the presence of the 1.3 extension block (byte 126=01) is forcing your system to treat the display as an HDTV (regardless of the label) and perhaps this is what's impacting on the output but in the case of this display, by default it's treated as a monitor. I did hope that forcing an HDTV timing might change the output behaviour but this doesn't appear to be the case so if we're looking at a bug it's very specific.

If you can think of a test you'd like me to try I'm more than happy to do so but so far I don't think I can confirm your findings.

Again, sorry.

Wo0zy

Edit: Bear with me. I may have found something

HT Slider

06-08-08, 11:29 AM

Not sure where to start.

Firstly this system is using 8.4 (will happily update it to 8.5 but I doubt it'll effect things). Under "Display Manager" by default I'm only offered the resolutions available in the EDID. If I go to the "Digital Panel" options and look under "HDTV Support", the bottom selection box offers no HDTV timings at all. If I tick an HDTV timing in the top box (to force the timing to be available in display manager) it does become selectable both in display manager and in the lower box in "HDTV Support". This TV doesn't like 1080i (shakes like hell) but 720p works fine. With 720p "forced" I can use either the full 1280x720 or "optimised" 1152x648 timing or even "add" a new timing to compensate for overscan (resulting in 1168x676 active pixels within a standard 720p timing). Regardless, selecting any of these timings has no impact on the color formats or greyscale levels with either the dongle in-line or using a straight DVI>DVI connection.

I'm not really sure about your definition of HDTV mode enabled? Like I said, by default I'm offered the resolutions found in the EDID (which are PC resolutions). If I then force an HDTV timing only the HDTV timings are offered in display manager until I again "force" a PC resolution then I'm once again offered the EDID timings. I guess this means it's switching between the two modes?

TBH I'm not really sure where this leaves things. I'm sure that, by default, this display is treated as a monitor but I can force HDTV timings if desired but neither scenario changes the output as far as I can see (except the resolution and therefore scaling).

It's likely that the presence of the 1.3 extension block (byte 126=01) is forcing your system to treat the display as an HDTV (regardless of the label) and perhaps this is what's impacting on the output but in the case of this display, by default it's treated as a monitor. I did hope that forcing an HDTV timing might change the output behaviour but this doesn't appear to be the case so if we're looking at a bug it's very specific.

If you can think of a test you'd like me to try I'm more than happy to do so but so far I don't think I can confirm your findings.

Again, sorry.

Wo0zy

Edit: Bear with me. I may have found something

Looking at the EDID, I'm not surprised that ATI interprets this LG HDTV to be a monitor. Running the EDID through your DTD Calculator, it includes a combination of standard PC resolutions and a couple of specific resolutions/timings for that particular display. I don't see any information in the EDID that would indicate it is an HDTV at all. It doesn't even have 720p nor 1080i in the EDID.

In contrast, my Toshiba doesn't have any standard PC resolutions at all (other than 640x480) and the preferred resolutions are all standard HDTV resolutions/formats. Looking at the EDID, it is easy to recognize the Toshiba as an HDTV (even without the EDID extension).

Is that LG one of those early combination HDTV/PC monitors that were essentially larger PC monitors with a built in TV tuner?

I'm not certain, but I doubt you'll see any difference between 8.4 and 8.5. For some reason I think ATI changed the HDTV detection between 8.3 and 8.4, but I still don't think the LG will be detected as an HDTV (based on the EDID).

I guess the only test that might be a little interesting is to see what the grey levels look like anyway. The ATI card will almost certainly send it RGB 0-255, since it thinks it is a PC monitor. The unknown is if the LG wants to receive RGB 0-255 or RGB 16-235. In this case (based on the EDID), I would say ATI is correct in sending it RGB 0-255.

My ATI issue, with my particular HDTV, is since it clearly is an HDTV and is recognized as such, that ATI should be sending it RGB 16-235 by default.

Wo0zy

06-08-08, 01:33 PM

Looking at the EDID, I'm not surprised that ATI interprets this LG HDTV to be a monitor. Running the EDID through your DTD Calculator, it includes a combination of standard PC resolutions and a couple of specific resolutions/timings for that particular display. I don't see any information in the EDID that would indicate it is an HDTV at all. It doesn't even have 720p nor 1080i in the EDID.

In contrast, my Toshiba doesn't have any standard PC resolutions at all (other than 640x480) and the preferred resolutions are all standard HDTV resolutions/formats. Looking at the EDID, it is easy to recognize the Toshiba as an HDTV (even without the EDID extension).

Is that LG one of those early combination HDTV/PC monitors that were essentially larger PC monitors with a built in TV tuner?

I'm not certain, but I doubt you'll see any difference between 8.4 and 8.5. For some reason I think ATI changed the HDTV detection between 8.3 and 8.4, but I still don't think the LG will be detected as an HDTV (based on the EDID).

I guess the only test that might be a little interesting is to see what the grey levels look like anyway. The ATI card will almost certainly send it RGB 0-255, since it thinks it is a PC monitor. The unknown is if the LG wants to receive RGB 0-255 or RGB 16-235. In this case (based on the EDID), I would say ATI is correct in sending it RGB 0-255.

My ATI issue, with my particular HDTV, is since it clearly is an HDTV and is recognized as such, that ATI should be sending it RGB 16-235 by default.

Fairly sure that's what I said;)

I'll repeat that it is possible to "force" standard HDTV 720p and that works fine. Obviously it needs forcing because there's nothing in the EDID to suggest it would work. There's no such thing as an "Established timing" for 720p, there's no DTD included in the EDID and without the 1.3 extension there's no SVD support. Still, 720p works and can be manipulated (I was even able to force 50Hz which I'm told never worked with this display using either NVIDIA or Intel graphics) and the timing then appears in ther HDTV section so sending an HDTV with the dongle connected isn't enough to force compression.

As Archibael mentioned before the appears to be nothing in EDID to determine whether RGB should be sent as 16-235 or 0-255 but as we're talking about a PC source, I would have to assume it sends 0-255.

It's a pity we can't find out exactly what the ATI drivers are looking for to force compression as the dongle's presence alone (and the fact that you're sending an HDTV timing) doesn't seem to be trigger it. Is it simply the presence of an extension block or something within the extension block?

I'm confident that with this TV/Monitor the same levels are sent with or without the dongle and regardless of whether I'm sending an HDTV timing or a PC resolution. I wish we could get a look at your extension block :(. I think it might just hold the answer.

I may "experiment" with Powerstrips EDID writing capabilities and see what happens.

By the way, my edit at the end of the last post turned out to be nothing. I discovered that it's possible to "crash" CCC if you try too many adjustments with VMC playing video. This resulted in CCC thinking that default brightness/contrast levels were set when they clearly weren't. Ending the process, relaunching CCC and applying defaults again fixed it :rolleyes:

Wo0zy

Jedd

06-08-08, 03:55 PM

Hello guys,
when driving an LCD projector (Sanyo Z2000 1080p) with 2600XT what refresh rate I'm supposed to set? By default it's 25Hz, but when panning or some action scenes things feel a little jerky, almost visually untraceable, but very irritating (my older 720p DSP has no problems like that at all)
If I switch to 24Hz I get a LOT of tearing. :confused:

Wo0zy

06-08-08, 06:52 PM

Dee_NA

06-09-08, 06:09 AM

I also used the ramp Arfster provided the link to. I also created a Media Center dvr-ms recording from it so I could test the calibration playing directly from Recorded TV in Media Center.

Once I used the HDMI adapter, UseBT601CSC=1 and cranked the "color" brightness/contrast to +31/73%, everything to do with Media Center and Media Player in SD and HD, including Arfster's recommended ramp and the dvr-ms ramp were all properly displaying 16-235 (and at the same time 0-255 for bitmaps). The only thing this didn't fix was PowerDVD, where I needed to calibrate it using separate Bluray/HD-DVD calibration DVDs (and ended up with the +19/-5 brightness/contrast).

BTW, the bmp and wmv ramps I attached in that earlier post were not created by me; they were created by Ian Kennedy a member of Microsoft's eHome (Media Center) software development team (TV playback quality group).

What's your thoughts of settings such as: brightness/contrast set at 50/100?
Wait til it goes to the TV once it's on a DVD or set before its burned?

If you wanna try somethin interesting, try using ContrastEnhancement in the registry... with application decide checked of course. You may notice deeper colors showing by doing so.

SaraJ

06-09-08, 07:30 AM

hey guys,

I've got 2600XT (the passive one from Gigabyte) installed in my HTPC on MCE 2005.
I can't get full screen. it just won't work, when I'm in full screen, all I see is a black, when I return to "windowed" mode (either for PowerDVD or Media Center), everything seems fine.
Any fix on this?

alecoddlyiv

06-09-08, 09:43 AM

I have a HIS HD 2600 pro connected via dvi to hdmi to a Samsung LE32R52BD which has a native resolution of 1366 x 768 and is I believe capable of 720p running on a 32bit vista ultimate HTPC.

Before the introduction of the graphics card the system was running at a resolution of 1366 x 768 via onboard vga without issue.

On installing the card the closest to native resolution is 1360 x 768 but the text on screen is a little pinched
It has detected a max resolution of 1920 x 1080 and this displays well but should I be running the card at this resolution for this tv?

When set to both 1360x and 1920x and I switch sources on the tv to anything else on going back to the screen the resolution has dropped to 1280x and no higher modes are available for selection without rebooting.

I created a profile through CCC and can use this to get back the option for 1920x but this is not ideal as it is a little too technical for my wife and virtually off the visible screen.

I have tried numerous drivers from manufacturers own through to new beta drivers from guru3d but the fault keeps occurring.

Heeellllllllp

Dee_NA

06-09-08, 10:13 AM

HT Slider

06-09-08, 10:52 AM

What's your thoughts of settings such as: brightness/contrast set at 50/100?
Wait til it goes to the TV once it's on a DVD or set before its burned?

If you wanna try somethin interesting, try using ContrastEnhancement in the registry... with application decide checked of course. You may notice deeper colors showing by doing so.

I'm not sure I understand your question.

I'm using CCC's "color" brightness/contrast to compress the 0-255 grey levels that the PC works with (and PC monitor's support) into 16-235 that is supported by my HDTV (all HDTVs, by default and viewing HD material at least, require 16-235; otherwise known as RGB BT.709).

Depending on the hookup and specific HDTV, the ATI cards sometimes do this conversion automatically and other times you need to adjust the brightness/contrast within the Catalyst Control Center.

If you are asking about arbitrarily setting the brightness/contrast to +50/100%, I very much doubt that will produce a decent image and if you then calibrated your HDTV for those levels, you would loose grey level resolution.

How do you use the registry setting for ContrastEnhancement? What does it do?

shurik_1

06-09-08, 02:05 PM

Hi, I do not know if anyonenoticed it but with latest PDVD 1730 patch I have at least 25% less CPU load with Blu-Rays wit Dolby TrueHD or DTS MA tracks. so now even my 4 years old Athlon XP 3000+ decodes them easily.

I also have a problem with latest ATI drivers in latest PDVD. Indeed, in HD DVD I get on the left side of the screen a black band. It happens with ATI 8.4 or 8.5 but not with 8.2. Does anyone have a clue?
thank you

Dee_NA

06-09-08, 03:50 PM

I'm not sure I understand your question.

I'm using CCC's "color" brightness/contrast to compress the 0-255 grey levels that the PC works with (and PC monitor's support) into 16-235 that is supported by my HDTV (all HDTVs, by default and viewing HD material at least, require 16-235; otherwise known as RGB BT.709).

Depending on the hookup and specific HDTV, the ATI cards sometimes do this conversion automatically and other times you need to adjust the brightness/contrast within the Catalyst Control Center.

If you are asking about arbitrarily setting the brightness/contrast to +50/100%, I very much doubt that will produce a decent image and if you then calibrated your HDTV for those levels, you would loose grey level resolution.

How do you use the registry setting for ContrastEnhancement? What does it do?

According to Ray Adams, programmer of ATi Tray Tools... ContrastEnhancement is a string value of "1" and suppose to go in the UMD/DXVA folder for those that have Vista. Not sure if it'll produce an image thats pleasing to the eye however as far as levels goes.

indieke2

06-09-08, 05:13 PM

Can't believe it! Installed the latest drivers to use total media theater with Ati.

Now my Ruby is not HDCP, neither is my Plasma panasonic!

Have desinstaled the latest drivers, tried the one's that worked, but after a few seconds the image freezes, because the HDCP is not recognized!

Tried everything, now I am tired

Wo0zy

06-09-08, 05:48 PM

Can anyone confirm that an Asus 2xxx/3xxx series card that shipped with a Rev.A (yellow) dongle still has working audio with the 8.4 or 8.5 drivers?

I've just finished testing two Asus cards. One shipped with the yellow dongle and one shipped with the grey dongle.

The card that shipped with the grey (Rev.b?) dongle provides audio over HDMI with 8.4 and 8.5 (providing the dongle is connected to the non-yellow DVI port), the card that shipped with the yellow (I assume Rev.a) dongle does not when connected to either port. The yellow dongle does however work fine with CCC 8.3 (and lower drivers) when connected to the yellow DVI port only :confused:

Please, someone save my sanity.

Considering the discussions we've been having here regarding the behaviour of the ATI dongles I'd love to know what hardware/software changes have taken place to cause this change. What exactly is different about the two revisions of dongle and what is CCC doing/looking for?

Any thoughts? I'm sure this is significant. Perhaps it's only the Asus cards that exhibit this problem? Perhaps it's all brands?

Any comments welcome.

Wo0zy

Well, I've just finished a few more tests and it seems obvious to me that Rev.A DVI>HDMI dongles DO NOT work with 8.4 or 8.5 but they do with 8.2. Even the Asus customised 8.493 driver doesn't work. The display is always seen as DVI not HDMI and therefore no audio.

Surely someone else has seen this????

I'm sure this is important. Why did ATI need a second revision of the dongle in the first place? What's changed in the drivers to break support for the REV.A dongles?

Anyone fancy having a guess?

Wo0zy

HT Slider

06-09-08, 11:24 PM

As Archibael mentioned before the appears to be nothing in EDID to determine whether RGB should be sent as 16-235 or 0-255 but as we're talking about a PC source, I would have to assume it sends 0-255.

I kept thinking there must be something somewhere that defines whether RGB 16-235 or RGB 0-255 should be used...

So I read through a most of the DVI, HDMI and EDID specifications and discovered section 6.6 in the "High-Definition Multimedia Interface Specification" (Version 1.2a) regarding "Video Quantization Ranges". Section 6.6 states:

6.6 Video Quantization Ranges

Black and white levels for video components shall be either "Full Range" or "Limited Range.” YCbCr components shall always be Limited Range while RGB components may be either Full Range or Limited Range. While using RGB, Limited Range shall be used for all video formats defined in EIA/CEA-861B, with the exception of VGA (640x480) format, which requires Full Range.

Section 6.6 also defines full range as 0-255 and limited range as 16-235. Also I dug through the EIA/CEA-861B specification (or the portions I managed to "find" on the Internet) and confirmed that both 720p and 1080i are defined within it.

To further define the colorspace, section 6.7 states that the BT.601 color space (16-235Y/240CbCr) is to be used for 480p, 480i, 576p, 576i, 240p and 288p and the BT.709 color space (16-235) is to be used for 1080p, 1080i and 720p.

Note, as previously determined, if the sink has an HDMI port, YCbCr is supposed to be output by the source. If the sink has a DVI port, RGB is supposed to be output by the source (i.e. section 6.6 does apply to HDTV's with DVI ports).

So - in other words, since ATI's video cards are certified to be HDMI spec compliant, they are supposed to output RGB 16-235 to an HDTV when 720p, 1080i or 1080p is the format in use and the HDTV is using a DVI port. By not doing this, ATI is in non-compliance of the HDMI specification.

I also dug through the next generation PC video connection spec, the "DisplayPort Standard" spec (Version 1, Revision 1a), and found the following:

5.1.1.1.1 RGB Colorimetry

All DisplayPort Source Devices must support RGB colorimetry with pixel depths of 18 and 24 bpp. Support
for 30, 36 and 48 bpp RGB is optional.

“VESA range” and “CEA range” are defined as follows:

“VESA range” must have:

Nominal zero intensity level at code value zero
Maximum intensity level at maximum code value allowed for bit depth.
i.e. 63 for 18 bpp RGB, 255 for 24 bpp RGB, 1023 for 30 bpp RGB, 4095 for 36 bpp RGB, and
65,535 for 48 bpp RGB.

“CEA range” must have:

Nominal zero intensity level at 16 for 24 bpp, 64 for 30 bpp, 256 for 36 bpp, and 4,096 for 48 bpp.
Maximum intensity level at maximum code value allowed for bit depth, namely, 235 for 24 bpp RGB, 940 for 30 bpp RGB, 3760 for 36 bpp RGB, and 60160 for 48 bpp RGB.
Note: The RGB CEA range is defined for 24, 30, 36, 48 bpp RGB only, not for 18-bpp RGB.

When a Source Device is transmitting a RGB stream with a video timing format called out in CEA-861C Section 5 as using CEA range RGB, it must use the CEA range RGB.

This DisplayPort spec clearly spells it out. RGB 0-255 for VESA/PC resolutions/formats and RGB 16-235 for CEA/HDTV resolutions/formats.

One additional and applicable spec requirement I noticed in several places is that HDTVs use a BT.709 color space for digital HD video (HDMI/DVI). YCbCr and RGB both support BT.709 and BT.709 defines the functional 8-bit range as 16 to 235. HDTVs can also support the full sRGB (0-255) colorspace, but I didn't see anything anywhere that requires this.

It's a pity we can't find out exactly what the ATI drivers are looking for to force compression as the dongle's presence alone (and the fact that you're sending an HDTV timing) doesn't seem to be trigger it. Is it simply the presence of an extension block or something within the extension block?

I'm confident that with this TV/Monitor the same levels are sent with or without the dongle and regardless of whether I'm sending an HDTV timing or a PC resolution. I wish we could get a look at your extension block :(. I think it might just hold the answer.

If you scan through my more recent posts you'll discover that using jong1's tip to look in the registry that I did manage to assemble my complete EDID. It contains the following:

00 ff ff ff ff ff ff 00 52 62 00 01 62 91 02 00
1f 0d 01 03 80 00 00 00 0e ee 91 a3 54 4c 99 26
0f 50 54 20 00 00 01 01 01 01 01 01 01 01 01 01
01 01 01 01 01 01 01 1d 80 18 71 1c 16 20 58 2c
25 00 00 00 00 00 00 9e 8c 0a d0 8a 20 e0 2d 10
10 3e 96 00 00 00 00 00 00 18 00 00 00 fc 00 54
6f 73 68 69 62 61 20 54 56 0a 20 20 00 00 00 fd
00 3b 3d 1f 2e 08 00 0a 20 20 20 20 20 20 01 e2
02 02 04 04 01 1d 00 72 51 d0 1e 20 6e 28 55 00
00 00 00 00 00 1e 8c 0a a0 14 51 f0 16 00 26 7c
43 00 00 00 00 00 00 98 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 de

I may "experiment" with Powerstrips EDID writing capabilities and see what happens.

By the way, my edit at the end of the last post turned out to be nothing. I discovered that it's possible to "crash" CCC if you try too many adjustments with VMC playing video. This resulted in CCC thinking that default brightness/contrast levels were set when they clearly weren't. Ending the process, relaunching CCC and applying defaults again fixed it :rolleyes:

Wo0zy

I too have managed to crash CCC and/or the video player once or twice when fiddling with settings too much and changing resolutions while playing video. I'm not sure what exactly I did that triggered it though. I found you can restart CCC through the start menu.

HT Slider

06-09-08, 11:41 PM

Well, I've just finished a few more tests and it seems obvious to me that Rev.A DVI>HDMI dongles DO NOT work with 8.4 or 8.5 but they do with 8.2. Even the Asus customised 8.493 driver doesn't work. The display is always seen as DVI not HDMI and therefore no audio.

Surely someone else has seen this????

I'm sure this is important. Why did ATI need a second revision of the dongle in the first place? What's changed in the drivers to break support for the REV.A dongles?

Anyone fancy having a guess?

Wo0zy

Very strange. No idea what is going on, but I see my new HIS 3870 includes a grey ATI HDMI dongle with the number 6140063500G on it.

The one that came with my genuine ATI HD 2600XT is black in color. A quick search on the web came up with HD 2600XTs as arriving with ATI HDMI dongle's with a part number 6141054300G.

From what I can tell (on the web), 6141054300G is for 2X00 cards and 6140063500G is for 3XX0 cards.

HT Slider

06-10-08, 12:22 AM

To make things even more clear, the "High-Definition Multimedia Interface Specification" (Version 1.2a) includes a list of resolutions/formats that fall under EIA/CEA-861B control and thus also fall under the requirement to be output using RGB 16-235 when the sink (HDTV) has a DVI port.

The HDMI specification can be downloaded from:
http://www.hdmi.org/download/HDMI_Specification_1.2a.pdf

Detailed timing is found in EIA/CEA-861B or a later version of CEA-861B for the following video format timings.

6.3.1 Primary Video Format Timings

640x480p @ 59.94/60Hz
1280x720p @ 59.94/60Hz
1920x1080i @ 59.94/60Hz
720x480p @ 59.94/60Hz
720(1440)x480i @ 59.94/60Hz
1280x720p @ 50Hz
1920x1080i @ 50Hz
720x576p @ 50Hz
720(1440)x576i @ 50Hz

6.3.2 Secondary Video Format Timings

720(1440)x240p @ 59.94/60Hz
2880x480i @ 59.94/60Hz
2880x240p @ 59.94/60Hz
1440x480p @ 59.94/60Hz
1920x1080p @ 59.94/60Hz
720(1440)x288p @ 50Hz
2880x576i @ 50Hz
2880x288p @ 50Hz
1440x576p @ 50Hz
1920x1080p @ 50Hz
1920x1080p @ 23.98/24Hz
1920x1080p @ 25Hz
1920x1080p @ 29.97/30Hz
2880x480p @ 59.94/60Hz
2880x576p @ 50Hz
1920x1080i (1250 total) @ 50Hz
720(1440)x480i @ 119.88/120Hz
720x480p @ 119.88/120Hz
1920x1080i @ 119.88/120Hz
1280x720p @ 119.88/120Hz

indieke2

06-10-08, 01:48 AM

I have posted this in another thread.

I have issues with HDCP. I really hate that thing! Ihave a Ruby and a Panasonic plasma. I know that when you put both displays on together HDCP will not work. I installed new drivers, the latest and suddenly HDCP on my Ruby would not work. Desinstalled reinstalled the old one's, nothing. Desinstalled, reinstalled original disc, then the drivers that worked, same, same.

Getting really tired of this I connected the computer to the A/V amplifier and then to the Plasma. It worked!

Then I restarted the computer, after putting my Ruby on the amplifier, and I did not even get an image on my display!

I would think, that maybe the 5 meter cable to the Ruby is guilty. But I guess not! Before installing the new drivers I had a REVERSE situation. The Ruby did work with Blu Ray, and it was the Panasonic who got the HDCP error message!

Anyway, whatever I do, it doesn't seem to change this situation.....Playing from the harddisc with my Ruby is possible, not my BD's!

Seems that Ati, is choosing the Panasonic for HDCP, but even when it is not connected, changing the output for the Ruby on the computer it doesn't work!

jong1

06-10-08, 08:31 AM

@HT Slider

Yes, you are right. We had commented on quite a bit of this earlier, including the specific quote from 6.,6. Not quite sure what you are concluding from it though.

Firstly, as I said in my post some days back, everything you mention is simply stated as part of the HDMI spec. There is no possibility of 'negotiation' between source and sink on black levels (e.g. via EDID or via the AVI Infoframe). The black level is mandated by the colorspace (and in one case resolution) being used. But anyway all of this only applies if the sink and source are both HDMI devices, which is not the case for you.

If the source (i.e the ATI card) determines a device is HDMI then the full HDMI spec must be followed. However, if the HDMI VSB is not there they device is treated as DVI and then the DVI spec is to be used, except for a few mostly low level things specified in C.1. C.2 "helpfully" gives some examples of what this means (including that RGB must be used), but these are not comprehensive. (By the way, I do not think the HDMI spec says that YCbCr must be used. Either RGB or YCbCr can be used. In fact HDMI devices do not HAVE to support YCbCr, although they must support it if they support it on any other input.)

So for your TV all the HDMI spec says is that:

- at a low level (electrical etc.) obviously it still has to use the HDMI interface
- other than that it has to communicate according to DVI 1.0.

Trouble is, as someone else posted the DVI spec is silent on colorspace and black levels (written largely before high quality video was a concern). For sure, there is no negotiation on black levels over DVI, or even recommendations. "They are what they are" as far as we can see.

So I think from a straightforward spec point of view ATI can do pretty much what they like within the DVI 1.0 spec :(. However, I agree it would be logical for them to treat a DVI HDTV as if it were an HDMI device in this respect or even better to allow a choice in CCC. The thing is, it seems to be trying to do just that - the dongle is compressing 0-255 into 16-235. The problem appears to be that are getting "double expansion" of video and single expansion even of "the desktop", inc. photos, before the dongle gets to work. If we could eliminate that then the dongle would be working fine for you I think.

digitlman

06-10-08, 09:23 AM

I have a 2600HD AGP version that works great, in xp and vista both with powerdvd and TMT (vista only) i have hardware accell working perfectly with all hd-dvd and bluray movies. i'm currently using vista with i think drivers from last november.

i setup a 2400HD PCIE version in vista for a friend. I installed the latest 8.5 drivers from ATI last night and both powerdvd and TMT are NOT using h/w acceleration. is there a reg setting or an older version of the driver i should try to get HW accell working?

usually people complain about the AGP version not accelerating, this is my first time messing with the pciE version. i expected it to work without issue.

with my agp card i installed jsut the drivers and the player and it works right away, i didnt have to do any wierd reg settings or anything.

also i run 720p output on my agp card, but the pciE 2400 card i am running 1080p output.

Wo0zy

06-10-08, 10:38 AM

goemonate

06-10-08, 11:06 AM

Wo0zy

06-10-08, 12:32 PM

recently got a Yamaha V863 receiver but it detects the HDMI signal from my HD2600 as DVI, therefore, there is no hdmi audio.... tried all driver from 7.13 to 8.5... no hope...

However, everything was fine with my last receiver onkyo 705, full hdmi picture and sound on all drivers.... weird enough?

Does thje REV.B dongle work on HD2600 card? if yes, does it fix the mis-detected to DVI problem? or should I just get HD3400 to fix the problem?

I'm sure I read a post from someone else having the same issue with a Yamaha receiver recently. IIRC it wasn't passing an EDID extension block (man it must seem like I'm obsessed with EDIDs :o).

Can you get a full EDID report using any of the tools discussed in the previous 3/4 pages? If so we can check to see if this is the issue. If it is, there may be a setting somewhere in the AVR's setup which causes this.

Do you have an HDMI TV? If so try connecting the HTPC directly to it just to confirm that the HDMI audio is still working after the mammoth driver experiment.

Cheers,

Wo0zy

jong1

06-10-08, 02:15 PM

The double expansion is indeed the issue as this isn't normal behaviour for either HDMI or DVI devices. But for me, equally odd is why the dongle has ANY impact on a connection recognised as DVI. It certainly doesn't for me :confused:

To be fair, I only have access to a Rev.A dongle but regardless of driver versions, it doesn't effect a DVI connection or an HDMI connection when it's recognised as DVI (see previous posts). More evidence for different behaviour with the Rev.B dongle?

Wo0zyWell I guess there is some logic to the dongle doing what it is doing as, according to HT, CCC is recognising his display as an "HDTV" even though it is DVI rather than HDMI.

But where that double expansion is coming from, that is a whole other matter! :confused:

HT Slider

06-10-08, 02:16 PM

So for your TV all the HDMI spec says is that:

- at a low level (electrical etc.) obviously it still has to use the HDMI interface
- other than that it has to communicate according to DVI 1.0.

Trouble is, as someone else posted the DVI spec is silent on colorspace and black levels (written largely before high quality video was a concern). For sure, there is no negotiation on black levels over DVI, or even recommendations. "They are what they are" as far as we can see.

In reading the HDMI spec, I didn't interpret that the HDMI spec was to be entirely ignored once a DVI device was connected.

Re-examining the HDMI 1.2a spec, I can see that in Appendix C it does essentially state that only the DVI 1.0 spec applies once a DVI sink is attached.

Even though all sources for HD content (including Cable, Satellite, Blu-ray, HD-DVD, etc.) use the BT.709 color space (16-235 or Limited Range RGB per the HDMI spec) and all HDTVs (by default) also use the BT.709 color space, there doesnt seem to be a spelled out, specific requirement for a source to actually output using BT.709 when the HDTV in use specifically has a DVI port on it.

There is one line in the HDMI spec that could potentially be argued to require BT.709 be used when an HDTV is hooked up through DVI, but it is vague at best. With the knowledge that all HD video is transmitted using the BT.709 color space and the knowledge that all HDTVs are by default calibrated to accept and display HD video that is using the BT.709 color space - consider the first sentence in the HDMI spec, Appendix C.2:

C.2 HDMI Source Requirements

When communicating with a DVI sink device, an HDMI Source shall operate in a mode compatible with that device.

It goes on to define some of the specific parameters (nothing stating BT.709 color space), but this statement could be interpreted as to expect the color space to also be compatible with the HDTV.

So I think from a straightforward spec point of view ATI can do pretty much what they like within the DVI 1.0 spec . However, I agree it would be logical for them to treat a DVI HDTV as if it were an HDMI device in this respect or even better to allow a choice in CCC. The thing is, it seems to be trying to do just that - the dongle is compressing 0-255 into 16-235. The problem appears to be that are getting "double expansion" of video and single expansion even of "the desktop", inc. photos, before the dongle gets to work. If we could eliminate that then the dongle would be working fine for you I think.

We don't know that the dongle is supposed to always compress 0-255 into 16-235, at least I don't.

What I know is when I use the HDMI dongle my HDTV "appears" to be receiving RGB 0-255 and when the HDMI dongle is not used my HDTV "appears" to be receiving a signal with an even greater level of expansion (approximating the transform required to convert 16-235 into 0-255).

My theory is ATI uses two different modes of operation depending on if the HDMI dongle is used or not. One, without the dongle, where they perform a DVI handshake and, if the sink is DVI, follow DVI rules and the other, with the HDMI dongle present, where they perform an HDMI handshake and, if the sink is HDMI, follow HDMI rules. When there is a mismatch with the dongle's use vs the sink device connector, they seem to default to RGB 0-255.

If this theory is correct, they appear to have a bug somewhere in their DVI to DVI setup and in some situations perform an expansion instead of a compression when a DVI HDTV is detected.

jong1

06-10-08, 02:39 PM

There is one line in the HDMI spec that could potentially be argued to require BT.709 be used when an HDTV is hooked up through DVI, but it is vague at best. With the knowledge that all HD video is transmitted using the BT.709 color space and the knowledge that all HDTVs are by default calibrated to accept and display HD video that is using the BT.709 color space - consider the first sentence in the HDMI spec, Appendix C.2:

It goes on to define some of the specific parameters (nothing stating BT.709 color space), but this statement could be interpreted as to expect the color space to also be compatible with the HDTV.BT.709 requires 16-235 levels, but so does BT.601 the SD colorspace. Trouble is that DVI is really a PC monitor spec written before high quality video on a PC was really considered, so DVI does not really concern itself with communicating colorspaces between the source and sink. It is up to the PC to convert the video into something suitable for the device based on its EDID. The most normal assumption would be RGB 0-255. I still don't know how the PC is supposed to know for sure it has a DVI HDTV connected, so 16-235 might be better; There is no clear flag (like the HDMI VSB) to tell it that. But I can see it might be reasonable to assume that if the device supports HDTV resolutions and not most PC resolutions.

That statement in C.2 is just rubbish isn't it! What is it supposed to mean? Surely they should say "communicate using DVI 1.0 (except as specified in C.1)" if that is what they mean, or else what do they mean?!

We don't know that the dongle is supposed to always compress 0-255 into 16-235, at least I don't.

What I know is when I use the HDMI dongle my HDTV "appears" to be receiving RGB 0-255 and when the HDMI dongle is not used my HDTV "appears" to be receiving a signal with an even greater level of expansion (approximating the transform required to convert 16-235 into 0-255).

My theory is ATI uses two different modes of operation depending on if the HDMI dongle is used or not. One, without the dongle, where they perform a DVI handshake and, if the sink is DVI, follow DVI rules and the other, with the HDMI dongle present, where they perform an HDMI handshake and, if the sink is HDMI, follow HDMI rules. When there is a mismatch with the dongle's use vs the sink device connector, they seem to default to RGB 0-255.

If this theory is correct, they appear to have a bug somewhere in their DVI to DVI setup and in some situations perform an expansion instead of a compression when a DVI HDTV is detected.Yeah your right, we don't know exactly how any of this is supposed to work, especially when different versions of the dongle behave differently with different driver versions! Your theory is quite reasonable too, maybe even better.

jasonbern

06-10-08, 04:40 PM

Hi,

This is my first post and i can be bad at writing English. I got an Sapphire 2400XT card and it is seems pretty ok if i use it at general tasks. But -like all we are, i was tried to use it to accelerate HD videos. I am using MPC Home Cinema with cyberlink h264 codecs and DXVA enabled. This setup always giving me macro pixelated images. Here is a sample...

http://img183.imageshack.us/my.php?image=pixelationae5.jpg

So what should i think? Does that mean my card is broken? Or should i think codecs are not ready to use?

arfster

06-10-08, 05:04 PM

jasonbern

06-10-08, 05:26 PM

If it's doing that with everything, the card is overheating. I had the same Sapphire card, did that until I directed a fan towards it.

Unfortunately, not a good card. I'd say just take it back.

No, this pixelation happening at just HD videos which i played with cyberlink codec. And it is only consuming %20 or less GPU power. I have played some games which uses %100 GPU power and they were completely ok.

And the images which produced by VLC or KMPlayer are ok. But HA is not working at vlc or kmp.

Today was the first day with this card. And now i am sooner to do your advice. Thanks for the reply.

HT Slider

06-10-08, 07:24 PM

BT.709 requires 16-235 levels, but so does BT.601 the SD colorspace. Trouble is that DVI is really a PC monitor spec written before high quality video on a PC was really considered, so DVI does not really concern itself with communicating colorspaces between the source and sink. It is up to the PC to convert the video into something suitable for the device based on its EDID. The most normal assumption would be RGB 0-255. I still don't know how the PC is supposed to know for sure it has a DVI HDTV connected, so 16-235 might be better; There is no clear flag (like the HDMI VSB) to tell it that. But I can see it might be reasonable to assume that if the device supports HDTV resolutions and not most PC resolutions.

I disagree with you. Although DVI was initially focused on PC monitors, it was specifically written to include support of digital high definition formats, including 720p, 1080i and 1080p. These formats use BT.709 for colorspace and always have.

By roughly 2001 much of the high end video equipment, including HDTVs, Projectors, DVRs, STBs, etc. had DVI ports on them. DVI was used on HDTVs and other video equipment until only a few years ago when they finally switched to HDMI. As far as I'm aware, DVI was the only means to transfer high definition video digitally from 1999 all the way up until 2004.

The HDMI spec, version 1.0 wasn't even released until the very end of 2003 and then device manufacturers had to start integrating HDMI into their video equipment. HDTVs must have been available well into late 2004, likely into 2005, with DVI ports on them instead of HDMI ports.

I also believe you will find most, if not all, audio/video experts would argue that when standard format (720p, 1080i, 1080p) HD video is transmitted over DVI that BT.709 is what is "normally" used, not, as you suggest, full range RGB 0-255. Similarly when standard format SD video is transmitted over DVI that BT.601 is "normally" used. BTW, BT.601 isn't exactly 16-235. It uses 16-235 for Y, but 16-240 for Cb and Cr.

While we agree that the specs poorly define the color space, we are going to have to agree to disagree on what is "normal" as far as full range RGB vs BT.709/BT.601 for transmission of video over DVI.

HT Slider

06-10-08, 07:41 PM

I still don't know how the PC is supposed to know for sure it has a DVI HDTV connected, so 16-235 might be better; There is no clear flag (like the HDMI VSB) to tell it that. But I can see it might be reasonable to assume that if the device supports HDTV resolutions and not most PC resolutions.

This whole DVI spec not mentioning BT.709 has got me thinking of another, much bigger feature of DVI that is also not mentioned in the spec anywhere - HDCP.

How can a "complete" DVI spec not mention HDCP anywhere within it, yet DVI is fully HDCP capable.

My HDTV for example is fully HDCP compliant.

You would almost expect a DVI spec 1.x or something to add mention of HDCP and similarly mention of the various color spaces in use.

jong1

06-11-08, 04:42 AM

While we agree that the specs poorly define the color space, we are going to have to agree to disagree on what is "normal" as far as full range RGB vs BT.709/BT.601 for transmission of video over DVI. I don't think there is much of a difference between us on this.

I certainly think CCC should allow people to select the option best for them instead of having the driver guess.

All I am saying is if the driver thinks it is connected to a PC monitor (even for 1080p) it is right and proper that, by default, it uses 0-255 levels. That is what people would want for non video applications like games and photo editing and if the monitor is calibrated for that all will be fine. I for one would not want the levels changing on my PC monitor when I switch between 1920x1200 and 1920x1080 for example.

However, in your case, the driver does seem to know it is connected to an HDTV so the point should be moot and it should be using 16-235 anyway.

jong1

06-11-08, 04:51 AM

This whole DVI spec not mentioning BT.709 has got me thinking of another, much bigger feature of DVI that is also not mentioned in the spec anywhere - HDCP.

How can a "complete" DVI spec not mention HDCP anywhere within it, yet DVI is fully HDCP capable.

My HDTV for example is fully HDCP compliant.

You would almost expect a DVI spec 1.x or something to add mention of HDCP and similarly mention of the various color spaces in use.I think this is because HDCP is a seperately licenced "add-on" to both HDMI and DVI. It is not "required" by either, although it is required for consumer equipment to be "HD Ready" (in the UK at least), so in practice HDMI normally means HDCP too.

It is barely mentioned in the HDMI spec, although as it is so commonly used there I think they felt they had to make some reference. Mostly they refer you to the HDCP specification.

curtis104

06-11-08, 10:43 AM

karrih

06-11-08, 10:50 AM

I have a ATI HD 2600XT that is running to my Mitsubushi HD1000 through a MonoPrice 4x1 HDMI switch, that seems to be having overscan issues using 1280 x 720 resolution. Anyone know of any solutions to this problem?

At least my Mits HC1100 lets me choose whether to have overscan or not, respective aspect ratio modes are called 16:9 and REAL, I believe. EDIT: it seems that also 16:9 has some overscan...

HT Slider

06-11-08, 01:04 PM

I have a ATI HD 2600XT that is running to my Mitsubushi HD1000 through a MonoPrice 4x1 HDMI switch, that seems to be having overscan issues using 1280 x 720 resolution. Anyone know of any solutions to this problem?

What is the problem? (please provide more detail)

Note that when using standard resolutions such as 1280x720 or 1920x1080 that ATI cards output spec compliant formats, just like a hardware STB, Blu-ray player, etc. This way the image (size wise anyway) looks exactly as intended by the display manufacturer. By design the amount of overscan is kept well within the "Safe Area" and only some of the "Invisible Area" is hidden by overscan.

This typically produces the best image, partially because the "Safe Area" is physically larger to the viewer, partially because HDTVs are typically optimized for spec compliant formats, and partially because this reduces the amount of scaling stages required. See: http://en.wikipedia.org/wiki/Title_safe and http://en.wikipedia.org/wiki/Overscan for more info.

gsr

06-11-08, 04:23 PM

What is the problem?
The problem would be that when using the display as a monitor for the computer, some important things (like the Windows Start button) end up in the overscan area. So overscan is fine when watching video, but not so great when trying to use the computer for other purposes.

HT Slider

06-11-08, 04:51 PM

The problem would be that when using the display as a monitor for the computer, some important things (like the Windows Start button) end up in the overscan area. So overscan is fine when watching video, but not so great when trying to use the computer for other purposes.

When using the PC as a PC, if you use any PC resolution or use a custom resolution overscan compensation should automatically be provided. It is only spec compliant 720p and 1080i/p resolutions (1280x720 and 1920x1080) that output spec compliant 720p/1080i/p formats.

All PC resolutions and custom resolutions are automatically compensated for overscan (as long as the video card recognizes your display as an HDTV with overscan and turns on overscan compensation).

If you are using Vista Media Center for example, configure Media Center to use 720p or 1080i/p and set the desktop resolution to whatever PC or custom resolution you prefer. When you start Media Center a spec compliant 720p or 1080i/p is automatically output and when you exit Media Center, the desktop's resolution will perfectly fill the screen, without any overscan.

This gives you the best of everything.

On the other hand, if for some reason you truly don't want to use spec compliant video output, you can instead tell Media Center to use any of the PC or custom resolutions and there will be no overscan. The choice is ultimately yours.

Note that this is the way it is supposed to work. I'm sure there are some HDTVs that are not detected correctly that are not driven this way by an ATI card.

jong1

06-11-08, 06:06 PM

Wo0zy

06-11-08, 06:45 PM

HT,

Save me going through pages of posts. Can you remind me what make/model your TV is?

Cheers,

Wo0zy

HT Slider

06-11-08, 06:47 PM

It's really easy in CCC ("under Scaling Options") to add underscan to your desktop which will still send 1920x1080 to your screen but will render the desktop into a window within that resolution, that just fills your screen.

It doesn't actually send 1920x1080 active pixels to the screen when you do this though (nor can any video card for that matter). The 1920x1080 is scaled into a smaller resolution, as you said in a window within the full 1920x1080, and that is sent to the display.

EDIT: If the HDTV supports variable pixel clock rates and variable total (not active) pixels, this may actually retain 1920x1080 pixels. Personally I have yet to see an HDTV that supports this and the slider is simply not there.

For viewing TV or movies, unless it was produced in the early 1970's or earlier (prior to "Safe Area" being used), the area lost due to overscan is not critical since the producers make sure nothing of value is included in this area. ATI's default of sending spec compliant HD video formats is generally a more appropriate way to view TV and movies and does not include overscan compensation for standard HDTV resolutions.

HT Slider

06-11-08, 06:48 PM

HT,

Save me going through pages of posts. Can you remind me what make/model your TV is?

Cheers,

Wo0zy

Toshiba 51H83

Wo0zy

06-11-08, 07:56 PM

It doesn't actually send 1920x1080 active pixels to the screen when you do this though (nor can any video card for that matter). The 1920x1080 is scaled into a smaller resolution, as you said in a window within the full 1920x1080, and that is sent to the display.

For viewing TV or movies, unless it was produced in the early 1970's or earlier (prior to "Safe Area" being used), the area lost due to overscan is only the "Invisible Area" and the producers make sure nothing of value is included in this area. ATI's default of sending spec compliant HD video formats (size wise at least) is a more appropriate way to view TV and movies and does not include overscan compensation for standard HDTV resolutions.

There's two different approaches to the resizing within CCC. The "scaling" slider and the "add" button within the HDTV support section. If you use the "add" button it actually sends the same HDTV timing as already selected but with less active pixels and a larger blanking interval. Exactly how the slider scaling works I don't know. Beyond what you've already mentioned (I assume you were talking about the scaling option) I've not heard a good technical description of how it's done. Anyone fancy taking a stab at it?

You can see the difference between the two methods when viewing an image with a sweeping curve (a couple of the standard Vista wallpapers are good).

Obviously for video, stick with the spec compliant timings as you say.

Wo0zy

jong1

06-12-08, 07:41 AM

It doesn't actually send 1920x1080 active pixels to the screen when you do this though (nor can any video card for that matter). The 1920x1080 is scaled into a smaller resolution, as you said in a window within the full 1920x1080, and that is sent to the display.

For viewing TV or movies, unless it was produced in the early 1970's or earlier (prior to "Safe Area" being used), the area lost due to overscan is only the "Invisible Area" and the producers make sure nothing of value is included in this area. ATI's default of sending spec compliant HD video formats (size wise at least) is a more appropriate way to view TV and movies and does not include overscan compensation for standard HDTV resolutions.Yeah, It is best to output full screen HD compliant video and display on a TV with no overscan. But he wants to see his taskbar and his TV always overscans, so for desktop work he needs overscan correction.

When it comes to displaying video on a TV I agree with you in principle but I am not sure how much difference it will make in practice. His TV is overscanning everything so he is never going to get bit perfect display. If he sends a full screen 720p image that is going to be overscanned by the TV say 5%+. If you send a 720p compliant picture but with overscan correction compressing the picture by the same amount beforehand, two scaling processes are taking place. This may look worse or it may be almost identical, neither is going to be bit perfect.

Personally I would try using the overscan correction (or set up a separate resolution as Wo0zy mentions); He will need that to get his desktop correctly displayed. Then compare video quality using this mode and normal overscanned 720p. If he is happy the quality is not significantly degraded it will obviously be simpler to stick with the overscan corrected mode. If not he can switch between the two modes for desktop and video work.

If he really wants to get clever he can use Reclock and Powerstrip to switch to a "full" 720p for video and back to "desktop" mode when the video finishes.

wally0206

06-12-08, 10:41 AM

arfster

06-12-08, 10:50 AM

Vista8.3 was best for me, anything later goes a bit nuts with dual screens, switching the taskbar at random, switching primary/secondary, changing resolutions at random *snore*

wally0206

06-12-08, 11:02 AM

arfster

06-12-08, 12:15 PM

thanks afster,

my main set-up is a dual display as well,I hate the switching primary/secondary on some of latest drivers, and they give me trouble with vc1 on my secondary display. I haven't gone back to Vista since mmmm....... 7.4 I think, do you like Vista or XP?

Some people hate it, but I've never had much of a problem with Vista. Unless you have a specific driver conflict I don't see any reason to use XP these days, and in future more and more stuff is being made with Vista in mind.

HT Slider

06-12-08, 05:39 PM

Yeah, It is best to output full screen HD compliant video and display on a TV with no overscan. But he wants to see his taskbar and his TV always overscans, so for desktop work he needs overscan correction.

When it comes to displaying video on a TV I agree with you in principle but I am not sure how much difference it will make in practice. His TV is overscanning everything so he is never going to get bit perfect display. If he sends a full screen 720p image that is going to be overscanned by the TV say 5%+. If you send a 720p compliant picture but with overscan correction compressing the picture by the same amount beforehand, two scaling processes are taking place. This may look worse or it may be almost identical, neither is going to be bit perfect.

Personally I would try using the overscan correction (or set up a separate resolution as Wo0zy mentions); He will need that to get his desktop correctly displayed. Then compare video quality using this mode and normal overscanned 720p. If he is happy the quality is not significantly degraded it will obviously be simpler to stick with the overscan corrected mode. If not he can switch between the two modes for desktop and video work.

If he really wants to get clever he can use Reclock and Powerstrip to switch to a "full" 720p for video and back to "desktop" mode when the video finishes.

You are making it much to complicated.

Simply using ATI's automatic and default HDTV mode is both the easiest way and best way to drive an HDTV with overscan. You don't need to use Reclock or Powerstrip at all. The video card will automatically switch to a full and compliant 720p or 1080i/p whenever 1280x720 or 1920x1080 is the selected resolution (so you simply configure your HTPC software to use one of these standard resolutions). At the same time, by selecting any standard PC resolution or one of the custom resolutions overscan compensation will automatically be applied. You don't need to turn on the overall overscan compensation, nor use Reclocd, nor use Powerstrip. This is all a part of what I consider ATI's HDTV mode.

Also I can assure you the core, "Safe Area" of the image will be clearer when displaying a spec compliant "full" 720p (1280x720) or 1080i/p (1920x1080) when compared to turning on the overall overscan compensation and shrinking both 720p and 1080i/p into a non-standard re-scaled format. In addition to fewer scaling operations (which "in theory" will affect the image quality), the core "Safe Area" portion of the image will be physically smaller. It is the size difference that makes the biggest difference to the details your eye can see from the couch. Looking at a persons face or the core area of a detailed scene it is very easy to see the difference.

To take advantage of ATI's automatic overscan compensation for PC and custom resolutions while still sending spec compliant 720p and 1080i/p for video applications here is what you need to do:

Install the regular Catalyst ATI driver with the regular Catalyst Control Center.
At this point, assuming your HDTV is automatically detected as an HDTV, if you use any standard PC resolution or the default "custom resolutions", it will automatically be underscanned by about 5% (likely with a thin black border around the screen). If you use any standard HDTV resolution (1280x720 or 1920x1080) it will be output using the full spec compliant format. For most casual users this is good enough and no further overscan adjustment is required.
If the ~5% isn't good enough for you and you want it perfect (no thin black borders or perhaps your display has more than 5% overscan), go into CCC, turn on Advanced controls (if not already enabled), select your display and get into the HDTV support section.
In here, at the bottom of the form, first select the standard format, such as "1080i/30 (1920x1080)", that you want to adjust the overscan for, click "apply format" and then click "Add". This allows you to create a "more optimized" resolution to use instead of the default ~5% "shrink".
Fine tune the size of the custom resolution so it perfectly fills your display, click OK and then apply.
At this point your new "custom resolution" will replace the original ~5% underscanned resolution that the card scales all standard PC resolutions into.
Do the same for the other format to get it perfect also (both 720p and 1080i/p) and if you use other formats (1080p/24, etc.), do the same for them also.
If you like you can delete the original 5% underscanned resolutions, but you don't need to. The ones you created will be used to scale all standard PC resolutions into anyway.

As a final step, make sure your HTPC software is configured to use a standard HDTV resolution (1280x720, 1920x1080, or whatever your HDTV prefers) and configure your desktop to use whatever standard PC resolution gives you the text size you like best. For video games, simply select whatever PC resolution your video card can handle.

Essentially the rule to follow is to use standard HDTV resolutions to output video and use standard PC resolutions (or the 1:1 pixel mapped custom resolutions you created) for PC applications and games.

My only beef with the way ATI has done this is they have not made fine tuning of the default overscan "intuitive" for the casual user. Functionality is excellent and about as good as it can be. For regular "non-enthusiast" users, I would like to see them create an "Overscan tuning wizard" where a click of a button starts the wizard and then the user simply adjusts the size of a box to perfectly fill the screen, followed by cliking OK (one "box" adjustment for each HDTV format).

Wo0zy

06-12-08, 06:00 PM

You are making it much to complicated.

Simply using ATI's automatic and default HDTV mode is both the easiest way and best way to drive an HDTV with overscan. You don't need to use Reclock or Powerstrip at all. The video card will automatically switch to a full and compliant 720p or 1080i/p whenever 1280x720 or 1920x1080 is the selected resolution (so you simply configure your HTPC software to use one of these standard resolutions). At the same time, by selecting any standard PC resolution or one of the custom resolutions overscan compensation will automatically be applied. You don't need to turn on the overall overscan compensation, nor use Reclocd, nor use Powerstrip. This is all a part of what I consider ATI's HDTV mode.

Also I can assure you the core, "Safe Area" of the image will be clearer when displaying a spec compliant "full" 720p (1280x720) or 1080i/p (1920x1080) when compared to turning on the overall overscan compensation and shrinking both 720p and 1080i/p into a non-standard re-scaled format. In addition to fewer scaling operations (which "in theory" will affect the image quality), the core "Safe Area" portion of the image will be physically smaller. It is the size difference that makes the biggest difference to the details your eye can see from the couch. Looking at a persons face or the core area of a detailed scene it is very easy to see the difference.

To take advantage of ATI's automatic overscan compensation for PC and custom resolutions while still sending spec compliant 720p and 1080i/p for video applications here is what you need to do:

Install the regular Catalyst ATI driver with the regular Catalyst Control Center.
At this point, assuming your HDTV is automatically detected as an HDTV, if you use any standard PC resolution or the default "custom resolutions", it will automatically be underscanned by about 5% (likely with a thin black border around the screen). If you use any standard HDTV resolution (1280x720 or 1920x1080) it will be output using the full spec compliant format. For most casual users this is good enough and no further overscan adjustment is required.
If the ~5% isn't good enough for you and you want it perfect (no thin black borders or perhaps your display has more than 5% overscan), go into CCC, turn on Advanced controls (if not already enabled), select your display and get into the HDTV support section.
In here, at the bottom of the form, first select the standard format, such as "1080i/30 (1920x1080)", that you want to adjust the overscan for, click "apply format" and then click "Add". This allows you to create a "more optimized" resolution to use instead of the default ~5% "shrink".
Fine tune the size of the custom resolution so it perfectly fills your display, click OK and then apply.
At this point your new "custom resolution" will replace the original ~5% underscanned resolution that the card scales all standard PC resolutions into.
Do the same for the other format to get it perfect also (both 720p and 1080i/p) and if you use other formats (1080p/24, etc.), do the same for them also.
If you like you can delete the original 5% underscanned resolutions, but you don't need to. The ones you created will be used to scale all standard PC resolutions into anyway.

As a final step, make sure your HTPC software is configured to use a standard HDTV resolution (1280x720, 1920x1080, or whatever your HDTV prefers) and configure your desktop to use whatever standard PC resolution gives you the text size you like best. For video games, simply select whatever PC resolution your video card can handle.

Essentially the rule to follow is to use standard HDTV resolutions to output video and use standard PC resolutions (or the 1:1 pixel mapped custom resolutions you created) for PC applications and games.

My only beef with the way ATI has done this is they have not made fine tuning of the default overscan "intuitive" for the casual user. Functionality is excellent and about as good as it can be. For regular "non-enthusiast" users, I would like to see them create an "Overscan tuning wizard" where a click of a button starts the wizard and then the user simply adjusts the size of a box to perfectly fill the screen, followed by cliking OK (one "box" adjustment for each HDTV format).

HT,

You didn't mention the CCC DTV "Scaling Options" slider? I thought that's what you were refering to in your previous post. I prefer the results using the method outlined in your last post but dragging a slider with overscan at one end and underscan at the other is fairly straight forward.

Although I don't like the results as much you could argue that using GPU scaling instead of physically adjusting the resolution in terms of active pixels/blanking interval may look better on some displays. It's also worth mentioning that more and more TVs have a "Just Scan" or "Exact Scan" or "Dot by Dot" setting which will compensate for overscan without any adjustment at the PC end.

I also meant to ask you, is your card from Asus? I made an interesting discovery about these today.

Wo0zy

HT Slider

06-12-08, 06:48 PM

HT,

You didn't mention the CCC DTV "Scaling Options" slider? I thought that's what you were refering to in your previous post. I prefer the results using the method outlined in your last post but dragging a slider with overscan at one end and underscan at the other is fairly straight forward.

The last time I tried the slider, the video card stopped being able to output spec compliant 720p and 1080i. Has ATI improved their "overall" Scaling Options slider so it doesn't affect standard HDTV formats or does it still apply to absolutely everything?

Actually I just went to the HTPC to give it a try and the overscan slider isn't there...

Is there something you have to do to enable it?

Although I don't like the results as much you could argue that using GPU scaling instead of physically adjusting the resolution in terms of active pixels/blanking interval may look better on some displays. It's also worth mentioning that more and more TVs have a "Just Scan" or "Exact Scan" or "Dot by Dot" setting which will compensate for overscan without any adjustment at the PC end.

I assume you mean using GPU scaling to scale one resolution into a window within another format/resolution vs playing with out of spec blanking intervals and scan rates. With my HDTV, it is very sensitive in the vertical (as most CRTs are). In the vertical I am better off to always stick with spec complaint numbers of vertical lines and blanking intervals. On the horizontal, I can increase the front and back porch to squish the image horizontally. Unfortunately, by doing this, the convergence goes off slightly and I need to fine tune it to compensate. Again, I prefer to simply send fully spec compliant formats and to simply view video this way (complete with the roughly 3-5% of overscan). Note I have calibrated the display through the service menu to slightly reduce the overscan though (again when DVI is used this essentially changes the CRT blanking intervals and again screws up the convergence if I push it too far).

These days I simply allow ATI's driver to always send spec compliant 720p/1080i and use the GPU scaling features to drive the display. Note that the method I outlined above does use GPU scaling for all PC resolutions. The custom resolutions are still spec compliant format wise, but as you mentioned the active pixels are reduced while the blanking interval is increased accordingly. Really this is essentially identical to only using the visible pixels and leaving the overscanned pixels as "black".

I also meant to ask you, is your card from Asus? I made an interesting discovery about these today.Wo0zy

No, my current card is a genuine ATI HD 2600XT. I do have an HIS 3870 IceQ3 sitting on top of the entertainment unit, ready to install. The only reason it isn't in yet is there are two hard drives I need to slightly move so they clear the massive IceQ3 cooler. I need to remove the particular hard drive mounting bracket and drill new holes in it...

We'll see how things go, but I "might" get around to the 3870 tonight. BTW, my HTPC case (http://www.digitalconnection.com/Products/Cases/ht400.asp), now a good 4+ years old, was originally designed to hold 3 hard drives and 1 DVD drive. I modified it years ago and created another 3 hard drive bays so it now has 5 hard drives (2.1 TB of storage in total) and 1 HD-DVD/Blu-ray combination drive installed.

jong1

06-12-08, 07:00 PM

You are making it much to complicated.

Simply using ATI's automatic and default HDTV mode is both the easiest way and best way to drive an HDTV with overscan. You don't need to use Reclock or Powerstrip at all. The video card will automatically switch to a full and compliant 720p or 1080i/p whenever 1280x720 or 1920x1080 is the selected resolution (so you simply configure your HTPC software to use one of these standard resolutions). At the same time, by selecting any standard PC resolution or one of the custom resolutions overscan compensation will automatically be applied. You don't need to turn on the overall overscan compensation, nor use Reclocd, nor use Powerstrip. This is all a part of what I consider ATI's HDTV mode.

Also I can assure you the core, "Safe Area" of the image will be clearer when displaying a spec compliant "full" 720p (1280x720) or 1080i/p (1920x1080) when compared to turning on the overall overscan compensation and shrinking both 720p and 1080i/p into a non-standard re-scaled format. In addition to fewer scaling operations (which "in theory" will affect the image quality), the core "Safe Area" portion of the image will be physically smaller. It is the size difference that makes the biggest difference to the details your eye can see from the couch. Looking at a persons face or the core area of a detailed scene it is very easy to see the difference.

To take advantage of ATI's automatic overscan compensation for PC and custom resolutions while still sending spec compliant 720p and 1080i/p for video applications here is what you need to do:

Install the regular Catalyst ATI driver with the regular Catalyst Control Center.
At this point, assuming your HDTV is automatically detected as an HDTV, if you use any standard PC resolution or the default "custom resolutions", it will automatically be underscanned by about 5% (likely with a thin black border around the screen). If you use any standard HDTV resolution (1280x720 or 1920x1080) it will be output using the full spec compliant format. For most casual users this is good enough and no further overscan adjustment is required.
If the ~5% isn't good enough for you and you want it perfect (no thin black borders or perhaps your display has more than 5% overscan), go into CCC, turn on Advanced controls (if not already enabled), select your display and get into the HDTV support section.
In here, at the bottom of the form, first select the standard format, such as "1080i/30 (1920x1080)", that you want to adjust the overscan for, click "apply format" and then click "Add". This allows you to create a "more optimized" resolution to use instead of the default ~5% "shrink".
Fine tune the size of the custom resolution so it perfectly fills your display, click OK and then apply.
At this point your new "custom resolution" will replace the original ~5% underscanned resolution that the card scales all standard PC resolutions into.
Do the same for the other format to get it perfect also (both 720p and 1080i/p) and if you use other formats (1080p/24, etc.), do the same for them also.
If you like you can delete the original 5% underscanned resolutions, but you don't need to. The ones you created will be used to scale all standard PC resolutions into anyway.

As a final step, make sure your HTPC software is configured to use a standard HDTV resolution (1280x720, 1920x1080, or whatever your HDTV prefers) and configure your desktop to use whatever standard PC resolution gives you the text size you like best. For video games, simply select whatever PC resolution your video card can handle.

Essentially the rule to follow is to use standard HDTV resolutions to output video and use standard PC resolutions (or the 1:1 pixel mapped custom resolutions you created) for PC applications and games.

My only beef with the way ATI has done this is they have not made fine tuning of the default overscan "intuitive" for the casual user. Functionality is excellent and about as good as it can be. For regular "non-enthusiast" users, I would like to see them create an "Overscan tuning wizard" where a click of a button starts the wizard and then the user simply adjusts the size of a box to perfectly fill the screen, followed by cliking OK (one "box" adjustment for each HDTV format).Now who is making it complicated!!

The guy said he had a problem where 720p was overscanned and he could not see the taskbar.

You said, to paraphrase, "get over it, what are you worried about". Then, when you accepted he had a need you suggested he switch between 2 resolutions for the different ways of working.

I said that given his screen is overscanning anyway he is not going to get "proper 720p" whatever he does. I said that if he uses the scaling options to make 720p fit, so he can see his taskbar, this may well, in practical terms, fix his problem. Depending on his screen he may well not be able to notice the difference between 720p scaled down by 5% (by his GPU) and then upscaled by 5% (by his TV) and just having it upscaled by 5% (by his TV), except he will see the whole image instead of around 95% of it (;)). So, in summary, set overscan correction and forget it. Can't get simpler than that.

I simply said, as an aside, that if he doesn't like the above and he feels he does wish to switch between underscanned 720p and full screen 720p for desktop and video work he can automate that, if he wishes, by using Reclock and Powerstrip.

Now you are suggesting he uses PC resolutions for desktop work and 720p for video. That is basically the same as option b above. You are suggesting his HTPC software switches to video mode whenever to loads it. OK. That works, but it is more complicated than my first option, you have to agree. Your option, if he can see problems with video using Scaling Options, avoids learning Reclock and Powerstrip, that is true and I do not underestimate that. But it is not the way I and many with HTPCs work. I do not use VMC. I do not expect to have to exit "HTPC mode" to use my desktop. I expect to be able to hit "Start" and bring up the taskbar at any time.

All that said I agree, what you suggest is a worthwhile alternative to Reclock and Powerstrip if he needs to work in 2 resolutions. However, still I say, the least complicated option is to;

a) try using Scaling Options to underscan 720p and leaving it at that for desktop and video

and failing that

b) Do what Wo0zy and you have suggested and create a custom resolution that corrects his overscan and leave it at that

Only if neither of those is satisfactory does he need to resort to different resolutions for video and desktop, which seems to be your suggested least complicated solution!

By the way, there is nothing "non-compliant" about the underscan performed by CCC. It still sends a compliant 720p image to the screen, simply with a reduced size desktop surrounded by a black border and as I have said before if he cannot turn off overscan on his TV he is never going to see a pixel perfect 720p image on his TV anyway.

p.s. I think the OP has wondered off in bemusement anyway! Well he did ask the question :)

Wo0zy

06-12-08, 07:39 PM

p.s. I think the OP has wondered off in bemusement anyway! Well he did ask the question :)

Guess it's lucky he doesn't have an Intel system in that case. DTD's anyone? :)

jong1

06-12-08, 08:00 PM

The last time I tried the slider, the video card stopped being able to output spec compliant 720p and 1080i. Has ATI improved their "overall" Scaling Options slider so it doesn't affect standard HDTV formats or does it still apply to absolutely everything?No. The scaling options are set per resolution.

My TV is 1080p and allows "Full pixel" operation (no overscan) for 1080p only. 720p is overscanned.

I tend to play games on my HTPC @720p - I am 9ft away and can barely tell the difference when running around like crazy in a FPS and it is not worth the performance penalty to go 1080p.

I have 720p set to underscan so the full desktop is visible, but 1080p is still output with zero underscan/distortion.

HT Slider

06-12-08, 09:49 PM

Now who is making it complicated!!
.
.
.
Now you are suggesting he uses PC resolutions for desktop work and 720p for video. That is basically the same as option b above. You are suggesting his HTPC software switches to video mode whenever to loads it. OK. That works, but it is more complicated than my first option, you have to agree. Your option, if he can see problems with video using Scaling Options, avoids learning Reclock and Powerstrip, that is true and I do not underestimate that. But it is not the way I and many with HTPCs work. I do not use VMC. I do not expect to have to exit "HTPC mode" to use my desktop. I expect to be able to hit "Start" and bring up the taskbar at any time.
.
.
.
By the way, there is nothing "non-compliant" about the underscan performed by CCC. It still sends a compliant 720p image to the screen, simply with a reduced size desktop surrounded by a black border and as I have said before if he cannot turn off overscan on his TV he is never going to see a pixel perfect 720p image on his TV anyway.

p.s. I think the OP has wondered off in bemusement anyway! Well he did ask the question :)

I stand by my statement that simply letting the video card automatically deal with overscan for PC resolutions and letting it output spec compliant video for HDTV resolutions is the simplest option. Also, if you change CCC settings to compensate for overscan when using 1280x720 or 1920x1080, you are no longer outputting spec compliant 720p and 1080i. There is no way around this fact.

I suggest you read up on HDTV video timings and video formats and then you will understand how it is impossible to change the overscan without diverging from the standard, spec compliant video format.

I don't know which HTPC front end software you are running, but I am running Vista Media Center. When Media Center is first set up, it runs through the TV setup wizard and here you select the native format for your HDTV (1080p or whatever your HDTV supports). Media Center automatically switches to this from then on, whenever you start it. I would assume that most, if not all, HTPC front ends would automatically switch to a standard HDTV resolution. As I said before, this then outputs a fully spec compliant resolution and format to the HDTV, just like an STB, HD-PVR, etc. would.

For the desktop and PC resolutions, as long as your HDTV reports within the EDID that it has overscan, the ATI card automatically compensates by producing an underscanned output so you can use the PC as a PC. Again, this is automatic - as long as your HDTV automatically triggers the overscanned HDTV mode for the ATI driver to run in.

To suggest that manually adjusting the overscan for each and every single resolution is simpler to set up than using the automatic overscan compensation simply doesn't make sense.

Does your system, by default, produce overscanned PC resolutions?

Since I've been running Vista, on every HDTV I've seen, including my "old" native 1080i set, ATI cards don't produce any overscan unless 1280x720 (720p) or 1920x1080 (1080i/p) is specifically used and then the overscan is simply the amount of overscan that the HDTV is designed/calibrated to produce when being driven with spec compliant HDTV formats.

HT Slider

06-12-08, 10:24 PM

No. The scaling options are set per resolution.

My TV is 1080p and allows "Full pixel" operation (no overscan) for 1080p only. 720p is overscanned.

I tend to play games on my HTPC @720p - I am 9ft away and can barely tell the difference when running around like crazy in a FPS and it is not worth the performance penalty to go 1080p.

I have 720p set to underscan so the full desktop is visible, but 1080p is still output with zero underscan/distortion.

My family and I do play games on our HTPC too, but we simply use the most appropriate (FPS wise) standard PC resolution when playing games. There are plenty to choose from and sometimes I manually select the 1:1 mapped custom resolution (although not very often to be honest).

We plug in the XBox 360 controller(s) and use the remote control to launch the game directly from within Media Center (with Vista, all games can be launched directly from Media Center with the remote control). The video game then automatically switches the video card to a standard PC resolution.

When we exit the game, Media Center pops back up and the video output reverts to spec compliant 1920x1080i again.

jong1

06-13-08, 02:13 AM

I stand by my statement that simply letting the video card automatically deal with overscan for PC resolutions and letting it output spec compliant video for HDTV resolutions is the simplest option. Also, if you change CCC settings to compensate for overscan when using 1280x720 or 1920x1080, you are no longer outputting spec compliant 720p and 1080i. There is no way around this fact.

I suggest you read up on HDTV video timings and video formats and then you will understand how it is impossible to change the overscan without diverging from the standard, spec compliant video format.

I don't know which HTPC front end software you are running, but I am running Vista Media Center. When Media Center is first set up, it runs through the TV setup wizard and here you select the native format for your HDTV (1080p or whatever your HDTV supports). Media Center automatically switches to this from then on, whenever you start it. I would assume that most, if not all, HTPC front ends would automatically switch to a standard HDTV resolution. As I said before, this then outputs a fully spec compliant resolution and format to the HDTV, just like an STB, HD-PVR, etc. would.

For the desktop and PC resolutions, as long as your HDTV reports within the EDID that it has overscan, the ATI card automatically compensates by producing an underscanned output so you can use the PC as a PC. Again, this is automatic - as long as your HDTV automatically triggers the overscanned HDTV mode for the ATI driver to run in.

To suggest that manually adjusting the overscan for each and every single resolution is simpler to set up than using the automatic overscan compensation simply doesn't make sense.

Does your system, by default, produce overscanned PC resolutions?

Since I've been running Vista, on every HDTV I've seen, including my "old" native 1080i set, ATI cards don't produce any overscan unless 1280x720 (720p) or 1920x1080 (1080i/p) is specifically used and then the overscan is simply the amount of overscan that the HDTV is designed/calibrated to produce when being driven with spec compliant HDTV formats.This risks boring everyone with a pointless point scoring operation! I think we are talking somewhat at cross purposes. You use VMC and I do not.

What you are saying works if he is using VMC and if he is happy to exit VMC to use the desktop. I use Meedio and that does not change the desktop resolution when you start it, although it can render over only a portion of the screen. I think this is pretty common for non VMC HTPC front ends. Of course it would be possible to emulate this functionality but for me it would be a backward step. I have Meedio set to render the full1920x1080 and my screen is in "full pixel" mode 1:1 mapped, so I do not have to do anything with overscan/underscan to use my desktop and watch video (all upscaled to 1080p) 1:1 mapped. Meedio hides the taskbar when it loads and whenever it is in focus, but all I have to do is hit "start" or "alt-tab" to another window and the taskbar appears, without changing resolutions. I do not want to have to exit Meedio when I want to use the desktop as i would need to with your solution if I had an overscanning TV. I also don't want to have to manually switch resolutions. Not all TVs can avoid overscan, as I said mine cannot @720p.

Using CCC's Scaling Options is just as easy as setting VMC's resolution and is HTPC front-end independent.

Your point about spec compliant video is irrelevant. The guy's TV is 1280x720 but it is overscanning, so even if you send 1280x720 he is actually getting something like 1152x648 pixels (just an example I do not know his actual overscan) upscaled to 1280x720. He will never see 1:1 pixel mapping so it is frankly irrelevant if what is being sent is "spec compliant" all that matters is if he can see the difference in PQ between 1152x648 ->(TV) 1280x720 v.s 1280x720 ->(PC) 1152x648 ->(TV) 1280x720. One thing he will notice is he will get more picture in the latter although I agree it may well be inessential detail. Now I don't know how each of these will look on his TV and you may be right, he may want to use 2 resolutions. If he does then your VMC option is a good way to go, if he (or anyone else reading this) uses VMC. However, again, a solution involving 2 resolutions which is independent of HTPC front-end would be more complex.

I don't understand your point about multiple adjustments to PC resolutions. How many resolutions do you run in? I use two - 1080p (95% of the time) and 720p (5% of the time, underscanned). The OP has a 720p set so he probably only needs one, if he is happy with the quality of the underscanned desktop, using one of the two methods.

Anyway, as I said at the start, we are talking at cross purposes somewhat. You are used to working in a different way to me that is fair enough, Windows offers so many options, that is its advantage and curse!

But that aside I do think working in one resolution for desktop and video is better than two, if video quality is not affected and whilst it is possible it might be, it is also possible it might not, when the image is being overscanned in either case.

I think we need to draw this to a close and not try to batter each other into submission. We just need to lay out the options and let the OP (if he has not lost the will to live) and anyone following in his footsteps try them out.

RussKingUK

06-13-08, 02:40 AM

Vista8.3 was best for me, anything later goes a bit nuts with dual screens, switching the taskbar at random, switching primary/secondary, changing resolutions at random *snore*

Hi Arfster,

8.4 would give me a strange stuttery picture so I have just downloaded 8.3 based on your comment. Are there any other hotfixes for this release that I must also install? What "must have" reg settings are there to use? Just the BT601 setting? I am running a 2600XT with a single display.

Cheers,

Russ.

Dee_NA

06-13-08, 06:21 AM

arfster

06-13-08, 07:51 AM

Hi Arfster,

8.4 would give me a strange stuttery picture so I have just downloaded 8.3 based on your comment. Are there any other hotfixes for this release that I must also install? What "must have" reg settings are there to use? Just the BT601 setting?

Yeah, just that is compulsory. If you're outputting to a HDTV, set CCC video to brightness16, contrast86 (usually not needed if you have the HDMI dongle though).

tman247

06-13-08, 08:27 AM

Just checking on what everyone thinks is the best working driver so far for your set-up?

Driver build#
XP or Vista
32bit or 64bit
pci express or agp

here is mine...

8.6beta
XP
32 Bit
pci express

I am thinking of doing a fresh intstall and going back to maybe 7.7

8.2
Vista
32bit
PCI-E (2600XT)

Was on 7.11 for ages prior to my recent system rebuild. I found 8.2 is almost perfect for all my HTPC needs, inc blu-ray/hd-dvd using PDVDU. I will be looking closely at the 4650 when it's out as a possible replacement (lower power, UVD2 etc)

tetsuo55

06-13-08, 09:38 AM

Cat 8.6 Beta version is available on www.guru3d.com I do believe for anyone that likes testing things. I haven't downloaded it myself yet since im still with 7.12; but please let me know if anyone does.

this version breaks DXVA and DirectX tests for me.
It did install on my 2400pro agp witouth hacks!

djos

06-13-08, 09:44 AM

this version breaks DXVA and DirectX tests for me.
It did install on my 2400pro agp witouth hacks!

Im not installing beta drivers on my VMC, the GF would kill me if they caused probs! :p

wally0206

06-13-08, 10:21 AM

I tried 8.6 beta, but I have gone back to 8.3 with a clean install. it seems the only driver that doesn't break DXVA on my seconday display. I have a little stutter with vc-1 still working on that.

arfster

06-13-08, 10:30 AM

I tried 8.6 beta, but I have gone back to 8.3 with a clean install. it seems the only driver that doesn't break DXVA on my seconday display. I have a little stutter with vc-1 still working on that.

Have you applied vforceuvdh264=1 & vforceuvdvc1=1? Dxva with two displays works fine for me with all drivers, once those are set.

wally0206

06-13-08, 10:54 AM

ok just found my mistake
I had powerdvd with show information checked, uncheck and no more stutter. man i am stupid.

HT Slider

06-13-08, 01:54 PM

The guy said he had a problem where 720p was overscanned and he could not see the taskbar.

You said, to paraphrase, "get over it, what are you worried about". Then, when you accepted he had a need you suggested he switch between 2 resolutions for the different ways of working.

He didn't say anything about the taskbar and I was simply asking what the problem was - not saying "get over it". There was nothing stated anywhere that suggested he was trying to get everything to work with only a single resolution in use.

Most people do use different resolutions for different games and applications (even if they don't realize it).

Your point about spec compliant video is irrelevant. The guy's TV is 1280x720 but it is overscanning, so even if you send 1280x720 he is actually getting something like 1152x648 pixels (just an example I do not know his actual overscan) upscaled to 1280x720. He will never see 1:1 pixel mapping so it is frankly irrelevant if what is being sent is "spec compliant" all that matters is if he can see the difference in PQ between 1152x648 ->(TV) 1280x720 v.s 1280x720 ->(PC) 1152x648 ->(TV) 1280x720. One thing he will notice is he will get more picture in the latter although I agree it may well be inessential detail. Now I don't know how each of these will look on his TV and you may be right, he may want to use 2 resolutions. If he does then your VMC option is a good way to go, if he (or anyone else reading this) uses VMC. However, again, a solution involving 2 resolutions which is independent of HTPC front-end would be more complex.

For someone who previously insisted that mapping 0-255 into 16-235 and back to 16-235 unacceptably looses color accuracy, your insistence on compromising video quality for the sake of not changing resolutions doesn't make much sense.

ATI cards do a very good job of removing overscan for PC resolutions while outputting spec compliant HDTV formats when HDTV resolutions are used so it makes sense to take advantage of this.

I will try one last time to explain how this works. Obviously you are free to do whatever you prefer.

To simplify things, consider a 720p HDTV with overscan with an actual viewing resolution of your suggested 1152x648. When a 720p signal is sent to it, out of the full 1280x720, only the center 1152x648 is visible. This is something you cannot change at the PC's end no matter what you do. When 720p is used, 1152x648 pixels are visible. If you use that HDTV as designed and feed it unaltered 720p, the center 1152x648 is displayed with a 1:1 pixel mapping and the area outside this is hidden. The "Action Safe Area" (http://en.wikipedia.org/wiki/Title_safe) is what matters and by driving the HDTV with a spec compliant 720p, this "Safe Area" is both large in size and clear.

If instead you underscan the output at the PC, the full 1280x720, including the unecessary "Invisible Area" is reduced into an 1152x648 resolution (more than a 20% reduction in bandwidth) and displayed. There is no longer a 1:1 pixel mapping and the important "Safe Area" is now much smaller in size. Read through the link mentioned above (http://en.wikipedia.org/wiki/Title_safe). By removing all of the overscan, you are viewing a portion of the image of little value to be viewed and you are doing so at the sacrifice of detail in the Action Safe Area.

Even if you consider your 1920x1080 display and intend to watch video at 720p, the same situation occurs (of course it would be better instead of use the PC to upconvert to 1080p). If your 720p has a visible 1152x648, your HDTV has essentially a map for each of those visible 1152x648 pixels onto your display's native resolution. If you take your full 1280x720 and scale it down into the 1152x648 you are once again reducing the size as well as using a lossy scaling operation. By lossy I mean you can never get the full image detail back. As you can see, it doesn't matter if the HDTV natively supports 720p or if it upconverts. There is still a virtual pixel mapping that only produces a reduced subset of visible pixels and that is all there are to work with. One you reduce the resolution, you can't get the detail back. I agree that in your situation that modifying your 720p doesn't matter since you are not using it for video. All I am saying is if you want to watch video using 720p, that leaving things configured for spec compliant 720p normally makes the most sense.

On top of that, we don't know anything about his HDTV. My HDTV supports both 720p and 1080i natively. His may support 720p natively too and may not even support 1080i/p.

Since modern HTPC software (such as Vista Media Center), automatically changes the resolution, there is no reason not to take advantage of the best image production capabilities that ATI has included with their video cards. Personally I stopped using Meedio about 4 years ago, but if I recall correctly I did have it configured to automatically switch to spec compliant 1920x1080 when it started.

I don't understand your point about multiple adjustments to PC resolutions. How many resolutions do you run in? I use two - 1080p (95% of the time) and 720p (5% of the time, underscanned). The OP has a 720p set so he probably only needs one, if he is happy with the quality of the underscanned desktop, using one of the two methods.

We use many different resolutions and I don't worry about it because all PC resolutions work perfectly. If my children install a new video game that defaults to 800x600, then that is what it typically stays at. Similarly if the game defaults to 1024x768, 1280x1024, 1360x768, or even 640x480 the game is more than often left with default settings. If I install a game, I do usually select a widescreen resolution if the game supports it. For Media Center, 1920x1080 is used and for the desktop 1360x768 is used (easy to read). All of this is seamless and automatic to the user.

To only use 720p for everything we would have to manually configure each and every game to use it. We have quite a few games that will only run using PC resolutions too so 720p isn't even available.

By using ATI's default overscan compensation, everything "just works". Since games don't defaults to 720p or 1080i/p, overscan with these compliant HDTV formats is never an issue.

Wo0zy

06-13-08, 02:55 PM

HT Slider

06-13-08, 06:12 PM

HT,

I'd still like to see you take a shot at explaining the difference between altering "spec compliant" timings by increasing the blanking interval and reducing the number of active pixels (which is what we do when we "add" a resolution" and therefore means that the timing is no longer fully compliant as defined in the EIA/CEA-861-B specification) as apposed to using the ATi card to scale the image within a fully spec compliant timing using the "scaling options" slider. I guess if your system doesn't offer this it's hard to experiment and see what I mean. Incidentally, the only time I've seen this is when the display is actually a monitor. Very odd considering the last few pages of posts :confused:

After that, perhaps we should start a new thread as these issues are not specific to the 2xxx series cards. Don't get me wrong, it's an interesting debate that I'm sure people will follow and I guess no one has complained but I'm just suggesting it out of respect.

Wo0zy

I don't know which method ATI is using for the "slider", but there are two different ways to reduce the effective size of an image by altering the timings (that I have used). This stuff is awkward to explain (well), especially since I don't discuss it much, but I'll try...

Rescale the image into a different resolution (smaller resolution to reduce overscan) and then display that resolution within a "window" using the standard format. This is definitely how ATI's HDTV overscan compensation works on my system as well as how their custom resolutions work on my system. Nvidia also uses this with their "overscan slider". This is the same as trading some active pixels for a longer blanking interval (keeping the overall pixel clock, overall scan lines and rates unchanged) and then scaling the original video into the reduced active pixel resolution - just a different way of describing it. When using Powerstrip, this is how the increase/decrease buttons work within the custom timing section if you lock the overall pixels, pixel clock and vertical refresh rate. The net result is that the image is smaller and a reduced resolution is used (reduced detail).

Increasing the overall (not active) number of horizontal and vertical pixels, pixel clock, blanking intervals (front and back porch in horizontal and vertical) and increasing the number of (blank) scan lines while leaving the active pixel count and the vertical refresh rate the same. This requires a monitor/display that is capable of a variable pixel clock, variable scan rates and variable display lines. Since HDTVs normally require specific video timings, this method doesn't normally work with these displays. It works well with a regular PC monitor though. When using Powerstrip, this is how the increase/decrease buttons work within the custom timing section by default (you may have to lock the active pixels, its been a while since I used it). The net result is that the image is smaller, but the same active resolution is used. Since greater image detail is retained initially, in theory this should produce a better image than #1 on most displays (if the display supports it).

To some extent, when using an analog connection (component), HDTVs can have the horizontal size reduced slightly by increasing the horizontal blanking interval and increasing the pixel clock while leaving the line count and vertical refresh rate unchanged. Using this method, the analog HDTV doesn't know that the pixel clock has increased. Since HDTVs typically can't handle any change in the number of scan lines, the vertical scale normally can't be changed this way, even with analog connections.

Note that if modern HDTVs are starting to support truly custom and variable video timings, including custom pixel clock rates (like a regular PC monitor) then it would be possible to control the overscan by #2 and this may be what the slider is doing.

Do you know of any way to read the pixel clock? If so, we could see if it changes with the slider.

One thing for certain: You cannot alter the amount of overscan by altering the timings without diverging from spec compliant video formats.

EDIT: Thinking about your comment "Incidentally, the only time I've seen this is when the display is actually a monitor. Very odd considering the last few pages of posts :confused: " a little more, I suspect the overscan slider only shows up on displays that do have the ability to vary the pixel clock, etc. and can perform #2 size adjustments.

gsr

06-13-08, 07:08 PM

To simplify things, consider a 720p HDTV with overscan with an actual viewing resolution of your suggested 1152x648. When a 720p signal is sent to it, out of the full 1280x720, only the center 1152x648 is visible. This is something you cannot change at the PC's end no matter what you do. When 720p is used, 1152x648 pixels are visible. If you use that HDTV as designed and feed it unaltered 720p, the center 1152x648 is displayed with a 1:1 pixel mapping and the area outside this is hidden.
If we're talking about fixed pixel displays (LCD, Plasma, DLP, etc.), are we assuming that the bezel on the TV is actually masking some pixels OR that the TV actually scales the image to create an overscan situation?

My dad's Toshiba LCD that he bought a few weeks ago has both a "native" mode and a "normal" mode (or something like that). The "native" mode does 1:1 pixel mapping, so if a computer is sending 1920x1080p to it, the entire Windows desktop is viewable. In the "normal" mode, that same image has overscan, so part of the desktop is chopped off on all 4 sides. One of those modes has to be scaling the image, and I'm assuming that the "normal" mode is the one doing the scaling. Personally, I'd prefer to avoid this sort of scaling as it's unlikely that it will make the picture look better (other than chopping off some stuff that we aren't really supposed to see in videos).

My ~6 year old Mitsubishi WS-65909 RPTV accepts 480i, 480p, 540p, and 1080i timings - anything else puts it into a very unhappy state. All of those modes have a non trivial amount of overscan. Some software that I use with the TV (such as Arcsoft TMT and PowerDVD) doesn't give me any way to adjust for the overscan, so if I go with the standard 1080i timings, I'm unable to see the control bars and some other features in these programs. To get around this, I created a mode in Powerstrip that outputs 1800x480p in 540p/1080i timings. This gives me just a tiny bit over overscan and I can now use PowerDVD. Unfortunately, Arcsoft TMT gets all confused about the aspect ratio at this resolution, so I'm stuck using TMT at 1080i and not being able to see the control bar. One of many reasons why I want to get a new display :).

HT Slider

06-14-08, 02:04 AM

If we're talking about fixed pixel displays (LCD, Plasma, DLP, etc.), are we assuming that the bezel on the TV is actually masking some pixels OR that the TV actually scales the image to create an overscan situation?

My dad's Toshiba LCD that he bought a few weeks ago has both a "native" mode and a "normal" mode (or something like that). The "native" mode does 1:1 pixel mapping, so if a computer is sending 1920x1080p to it, the entire Windows desktop is viewable. In the "normal" mode, that same image has overscan, so part of the desktop is chopped off on all 4 sides. One of those modes has to be scaling the image, and I'm assuming that the "normal" mode is the one doing the scaling. Personally, I'd prefer to avoid this sort of scaling as it's unlikely that it will make the picture look better (other than chopping off some stuff that we aren't really supposed to see in videos).

My ~6 year old Mitsubishi WS-65909 RPTV accepts 480i, 480p, 540p, and 1080i timings - anything else puts it into a very unhappy state. All of those modes have a non trivial amount of overscan. Some software that I use with the TV (such as Arcsoft TMT and PowerDVD) doesn't give me any way to adjust for the overscan, so if I go with the standard 1080i timings, I'm unable to see the control bars and some other features in these programs. To get around this, I created a mode in Powerstrip that outputs 1800x480p in 540p/1080i timings. This gives me just a tiny bit over overscan and I can now use PowerDVD. Unfortunately, Arcsoft TMT gets all confused about the aspect ratio at this resolution, so I'm stuck using TMT at 1080i and not being able to see the control bar. One of many reasons why I want to get a new display :).

With modern displays, I'm confident that there is not a significant number of pixels actually hidden by the bezel; it is more of a situation where they are "virtually hidden by the bezel". My explanation is based on a display that does not have 1280x720 available "pixels", but instead has a slightly smaller subset of pixels (this is typical of a the 720p displays I've worked with)

I honestly don't know if the very latest 1080p and 720p displays have fewer than 1920x1080 and 1280x720 physical pixels, but older ones certainly did. With my HDTV, when 720p is used, only 704 horizontal lines are visible for example. This is still a 1:1 mapping, but everything beyond the center 704 is not visible.

I suspect that a native 1080p HDTV with a "normal" 1080p mode that overscans could have fewer than 1920x1080 actual pixels (just like 720p and 1080i displays). I may be wrong, but I don't believe there is a requirement for a native 1080p HDTV to actually have 1920x1080 pixels. As long as it is close enough that a 1:1 pixel map still displays the "Active Safe Area", it can still be sold as a native 1080p display. In other words, a 1080p display may only have as few as 1776x1000 visible/actual pixels. If someone can provide a link to any specifications that might clear this up; I would be interested in knowing for certain.

On the other hand, there are quite a few new 1080p native HDTVs where the "normal" 1080p mode does not have any overscan at all. One question I keep wondering about is: Does this mean that these particular displays actually have 1920x1080 physical and visible pixels while the others with "normal" 1080p overscan do not?

This is really getting off topic, but all of this unfortunately has a huge impact on how successful our video cards are at compensating for overscan without intervention by the user.

I have spoken to Cyberlink about the fact that PowerDVD Ultra is poorly designed for a display with overscan. They told me simply that PowerDVD Ultra is intended to be used with a PC monitor, not an HDTV (they even recommend standard PC resolutions be used, not 1920x1080 for some reason). Software that is truly intended to be used with an HDTV, such as Media Center, is designed in such away that all of the GUI is located within the "Title Safe Area".

With our system for video playback I always use a spec compliant 1920x1080i output, complete with overscan (on my HDTV). This means that PowerDVD does have some of the GUI hidden by the overscan, but I don't find this causes any real problems since it is launched directly from Media Center and is 100% controlled using the Media Center remote. With this setup, there isn't any need to see any of the PowerDVD GUI bits that are hidden.

One thing to note is Cyberlink does make a rehashed version of PowerDVD Ultra that is intended to be used with an HDTV. It is only sold to OEMS for their Media Center PCs. With it, the GUI is well within the "Title Safe Area". My father's Acer Media Center PC has this installed on it and it also fully integrates with Media Center. Cyberlink doesn't call the software PowerDVD Ultra, but calls it something like HD Arcade Deluxe. Updates for it seem to be available around the same time as PowerDVD Ultra updates and it is also able to play the same Blu-ray and HD-DVD movies as PowerDVD Ultra.

jong1

06-14-08, 05:58 AM

I will try one last time to explain how this works.I would appreciate a little less of the the arrogant superior tone especially when you don't know your facts.

First, you clearly do not know what 1:1 pixel mapping means. This means each pixel in the original video is mapped to one single pixel in the display (hence "1:1"). Clearly this is not possible if 1152x648 is being zoomed to fill 1280x720 actual pixels. Any deviation from this is damaging the PQ.

Second, with movies (where for the majority people picture quality is most important) there is no such thing as a part of the final print that "isn't even supposed to be displayed at all". The link you supplied is for conventional video production which is very different. Yes, the filmmaker needs to take into account that even in the cinema some overscan/cropping of the image might occur and shoot accordingly but that is a very different thing. He wants the whole 16:9 frame shown, he merely has to accept that reproduction equipment in the theatre is not always perfect.

Let's look at Blu-ray a bit more - 1920x1080 on disc (actually 1920x1088, for reasons of codec efficiency, but the last 8 pixels are simply cropped). The authors of the disc have gone to a great deal of trouble to get these 1920x1080 pixels as close to the original print as humanly possible. No owner of a 1080p display in their right mind, if they can avoid it, would destroy 1:1 pixel mapping by using overscan to zoom in on the safe area as you would appear to prefer.

Look we really do need to draw a line under this. Believe me I have always understood the implications and the possible impact on image quality of scaling the image on the PC (even if you do not understand the terminology quite as well as you would like to think). For the record, I have never ruled out your method or disagreed that in principle it offers the potential for higher picture quality. I am not underestimating the importance of PQ. I use a truly 1:1 pixel mapped 1080p display with no overscan and would not personally have anything less.

But you need to be more open to possible alternatives to your own prescriptive viewpoint. Particularly when you do not know the software or hardware environment that people may be using, which may impact the appropriateness of your preferred solution. You also need to learn to be less supercilious especially when you are not in complete mastery of your own facts.

jong1

06-14-08, 06:15 AM

Wo0zy

06-14-08, 08:01 AM

Do you know of any way to read the pixel clock? If so, we could see if it changes with the slider.

You can check what pixel clock the display requires from (here I go again:rolleyes:) the EDID. If a timng is offered in an SVD the pixel clock is defined by the EIA/CEA-861-B specification. If a timing is defined by a DTD the first 2 bytes represent the pixel clock for that timing.

You can use Powerstrip to see what pixel clock is being used. However, no adjustment to the pixel clock has to happen if the Total (not active) number of H and V pixels and the refresh rate isn't changed so I doubt it would prove anything.

This is what I believe happens when you "add" a resolution (and is definitely what happens when we adjust a timing using DTDCalc and Intel Graphics).

1. The Total number of H & V pixels remains the same
2. The H & V Sync Pulse widths remain the same
3. The refresh rate remains the same
4. Because of #1 and #3 the Pixel Clock remains the same
5. The number of "Active Pixels" (H and/or V) are reduced by a sufficient value to remove overscan.
6. The Front Porch and Back Porch values are increased (usually by the same value) to keep the total number of pixels the same.

In some cases, displays do odd things with timings that are adjusted in this manner which is typically caused by the displays internal scaler. Sometimes the values can be further adjusted (which may or may not result ia a change in the pixel clock) to compensate. Sometimes they can't. With Intel Graphics this means there are some rear cases when overscan compensation using the DTD method just doesn't work. With ATI (as Jong1 says) you can always use the mysterious "Scaling" slider instead (in most cases anyway ;)).

Even though things are getting a little "heated", I have enjoyed our three way discussions but I really don't think it's fair to continue them here so I'll make this my last "non-2xxx" specific post. PM me if you fancy starting another thread to continue things.

Cheers guys.

Wo0zy

leeperry

06-14-08, 12:39 PM

Just checking on what everyone thinks is the best working driver so far for your set-up?

Driver build#
XP or Vista
32bit or 64bit
pci express or agp

here is mine...

8.6beta
XP
32 Bit
pci express

I am thinking of doing a fresh intstall and going back to maybe 7.7

since the 7.12, the tasktray red icon dies randomly after 1/2 days.....very annoying as pstrip won't work if CCC.exe is not running

I know I'm not the only one with this problem....

this problem never occurs with the 7.11 crysis drivers....might just revert to this once for all.

I got a HD2600 on XP SP2, and even a clean reinstall won't fix this problem :mad:

still have the option to run the 8.x drivers + 7.11 CCC.....but since the 8.6 beta it doesn't seem to work anymore :(

HT Slider

06-14-08, 02:13 PM

I would appreciate a little less of the the arrogant superior tone especially when you don't know your facts.

I was the one trying to help someone and you started flaming me and accusing me of not knowing my facts. Take a look at where this started to see who had the arrogant attitude. Even here, you are blatantly insisting I don't know my facts. Arrogant, superior tone, no?

First, you clearly do not know what 1:1 pixel mapping means. This means each pixel in the original video is mapped to one single pixel in the display (hence "1:1"). Clearly this is not possible if 1152x648 is being zoomed to fill 1280x720 actual pixels. Any deviation from this is damaging the PQ.

How do you infer I don't know what 1:1 pixel mapping is? Where is your 1152x648 being "zoomed to fill 1280x720"? What are you talking about?

I have tried to be polite, but I suggest you read up on video timings first if you want to continue to flame me.

I will try to explain this one more time: If a native 720p display produces an overscanned image where only 1152x648 is visible, the image is being CROPPED, not zoomed. By sending a spec compliant full 1280x720p to that HDTV, the visible 1152x648 is being displayed without any zooming, scaling or anything and this will produce the best image quality for the "Active Safe Area". In addition to this, with most native 720p displays this will produce a 1:1 pixel mapping, even though, there may be overscan.

Why you keep insisting there is some sort of 1152x648 to 1280x720 scaling going on when a spec compliant format is used I will never know.

Second, with movies (where for the majority people picture quality is most important) there is no such thing as a part of the final print that "isn't even supposed to be displayed at all". The link you supplied is for conventional video production which is very different. Yes, the filmmaker needs to take into account that even in the cinema some overscan/cropping of the image might occur and shoot accordingly but that is a very different thing. He wants the whole 16:9 frame shown, he merely has to accept that reproduction equipment in the theatre is not always perfect.

If you talk to any movie producer, you will find they always take into account the "Action Safe Area" and this is for reproduction in movie theaters as well as TVs. Due to the variable cropping that takes place, they make sure nothing important is within the "future overscanned area" and they don't expect all viewers to be able to see this area.

I have provided links to back up this fact. If you want to insist that movies are not produced with an "Active Safe Area" considered for overscan, please back up your claims with a factual link.

All I have been saying all along is we are better off to output spec compliant video formats (720p, 1080i/p) and not worry about a small amount of overscan. Rescaling the image into a smaller resolution and thus producing a smaller image on the display only degrades the image quality.

From what I understand, you don't currently and never have owned an HDTV that produces an overscanned image with the native source format sent unaltered to the display - yet you keep insisting you are an expert on this. My HDTV natively supports both 720p and 1080i and both have on the order of 3-5% overscan. I have tested both fully spec compliant formats as well as custom underscanned resolutions.

No owner of a 1080p display in their right mind, if they can avoid it, would destroy 1:1 pixel mapping by using overscan to zoom in on the safe area as you would appear to prefer.

What are you talking about? Not once have I ever suggested rescaling the image to add overscan.

Look we really do need to draw a line under this. Believe me I have always understood the implications and the possible impact on image quality of scaling the image on the PC (even if you do not understand the terminology quite as well as you would like to think). For the record, I have never ruled out your method or disagreed that in principle it offers the potential for higher picture quality. I am not underestimating the importance of PQ. I use a truly 1:1 pixel mapped 1080p display with no overscan and would not personally have anything less.

But you need to be more open to possible alternatives to your own prescriptive viewpoint. Particularly when you do not know the software or hardware environment that people may be using, which may impact the appropriateness of your preferred solution. You also need to learn to be less supercilious especially when you are not in complete mastery of your own facts.

I agree we need to draw the line, however, you need to consider your own advice. While reading your post I kept laughing at how you insist I am being rude, yet keep interjecting with additional outrageous jabs.

My advice to the original poster, was that it is OK to have overscan and that spec compliant video formats typically produce the best image quality. You disagreed with me, then kept insisting that modifying the output to get rid of overscan still produces spec compliant 720p, then started going on about some artificial 1152x648 to 1280x720 upscaling that will loose image quality if you don't underscan the output, and on and on - yet now you are saying I need to be more open. Then to top it off you still insist that I don't know my "own facts". Please - grow up a little.

topcaser

06-14-08, 02:58 PM

Do you know this nice ATI BIOS Editor for 2x00 and 3x00 cards?
Click (http://www.aledander.de/vb/ati_bios_editor/support/index_eng.htm)

trazalca

06-15-08, 05:12 AM

red5goahead

06-15-08, 06:11 AM

hi all.

Let me know this strange behaviour of my Ati 3850. Last catalyst 8.5 installed
I'm Under Vista Bussiness with Sp1 installed.

I've this mpeg file (http://www.4shared.com/file/51371561/14f31b85/PIANETA_MARE_-_Rete_4_2008-06-15_11-10-50.html?dirPwdVerified=f1c1905)on 4shared.com free for the test (8 Mbyte only)

I open this file with Media Player Classic Home Cinema with this filtergraph
(dxva oriented)

http://img177.imageshack.us/img177/9897/21302487jc0.th.jpg (http://img177.imageshack.us/my.php?image=21302487jc0.jpg)

I check UseBT601CSC"="1" registry value in the registry. it's correct.

But the image in not expanded to evr in pc levels. it have a level to 16-235
(note the letterbox bands). so the image don't have the correct dark level and is not very good.

http://img292.imageshack.us/img292/9337/46964983ku8.th.jpg (http://img292.imageshack.us/my.php?image=46964983ku8.jpg)

Note If I use evr customer presentation in the Media Player Classi Home Cinema the image is correct to pc level. In Media Portal is the same video level 16-235 with its evr custom. MP is my main player.
Can you help me to find out the solution?

edit: I'm sorry. I had an Ati 2600 PRO and now a 3850. there is 2 entry in registry 0000 and 0001. the last one 0001 is related to 3850

Device Description=ATI Radeon HD 3800 Series

this node DO not had UseBT601CSC!!

arfster

06-15-08, 08:14 AM

i have some problems with my ati hd2600 pro with vista sp1 ccc 8.5, with dxva checker i found that no dxva is available for h264 avivo, (like photo) i use mpc hc in EVR mode and no external filters, with mkv file i can't have hardware acceleration. Someone is so gentle to tell me how can i get that? Thanks in advance

That photo is normal, you have dxvah264nofgt.

Most 10800 mkvs don't accelerate because the person who encoded them ignored the standards.

arfster

06-15-08, 08:15 AM

I check UseBT601CSC"="1" registry value in the registry. it's correct.

Best to use the exdeus script for this, there are numerous places in the registry that look similar.

leeperry

06-15-08, 09:40 AM

Do you know this nice ATI BIOS Editor for 2x00 and 3x00 cards?
Click (http://www.aledander.de/vb/ati_bios_editor/support/index_eng.htm)

cool! last time I checked it was only for 3xx0 :D

trazalca

06-15-08, 10:58 AM

Thanks arfster, but can you explain me the difference between the two??

topcaser

06-15-08, 11:59 AM

cool! last time I checked it was only for 3xx0 :D

Managed to get 42° GPU temp. watching SD Content by lowering voltage. Coming from 47°.

leeperry

06-15-08, 04:36 PM

don't lower the voltage too much, my HD2600 had a stock buggy BIOS that set 1.2V all the time, except for 2D they put 1.05V

it was giving black screens in windows.

Sapphire has given me a new BIOS update, that leaves it at 1.25V all the time.....no more problems :)

I was more thinking into going to 1.3V or sumthing, considering I've got a VF700-CU onboard :D

arfster

06-15-08, 04:57 PM

jong1

06-15-08, 05:20 PM

@HT Silider.

Great. Finally the discussion has run itself into the ground and we can move on.

Jon

Fuz

06-16-08, 12:32 AM

No owner of a 1080p display in their right mind, if they can avoid it, would destroy 1:1 pixel mapping by using overscan to zoom in on the safe area as you would appear to prefer.

I couldn't agree more. I hope overscan becomes a thing of the past in the near future, as there is no real need for it any more with digital media. There is no more "junk" on the sides of the screen that needs to be hidden as if we were watching movie on VHS.

I can't stress this enough... if you are in the market for a new HDTV, make sure there is an option to disable overscan.

Dee_NA

06-16-08, 01:48 AM

I was the one trying to help someone and you started flaming me and accusing me of not knowing my facts. Take a look at where this started to see who had the arrogant attitude. Even here, you are blatantly insisting I don't know my facts. Arrogant, superior tone, no?

How do you infer I don't know what 1:1 pixel mapping is? Where is your 1152x648 being "zoomed to fill 1280x720"? What are you talking about?

I have tried to be polite, but I suggest you read up on video timings first if you want to continue to flame me.

I will try to explain this one more time: If a native 720p display produces an overscanned image where only 1152x648 is visible, the image is being CROPPED, not zoomed. By sending a spec compliant full 1280x720p to that HDTV, the visible 1152x648 is being displayed without any zooming, scaling or anything and this will produce the best image quality for the "Active Safe Area". In addition to this, with most native 720p displays this will produce a 1:1 pixel mapping, even though, there may be overscan.

Why you keep insisting there is some sort of 1152x648 to 1280x720 scaling going on when a spec compliant format is used I will never know.

If you talk to any movie producer, you will find they always take into account the "Action Safe Area" and this is for reproduction in movie theaters as well as TVs. Due to the variable cropping that takes place, they make sure nothing important is within the "future overscanned area" and they don't expect all viewers to be able to see this area.

I have provided links to back up this fact. If you want to insist that movies are not produced with an "Active Safe Area" considered for overscan, please back up your claims with a factual link.

All I have been saying all along is we are better off to output spec compliant video formats (720p, 1080i/p) and not worry about a small amount of overscan. Rescaling the image into a smaller resolution and thus producing a smaller image on the display only degrades the image quality.

From what I understand, you don't currently and never have owned an HDTV that produces an overscanned image with the native source format sent unaltered to the display - yet you keep insisting you are an expert on this. My HDTV natively supports both 720p and 1080i and both have on the order of 3-5% overscan. I have tested both fully spec compliant formats as well as custom underscanned resolutions.

What are you talking about? Not once have I ever suggested rescaling the image to add overscan.

I agree we need to draw the line, however, you need to consider your own advice. While reading your post I kept laughing at how you insist I am being rude, yet keep interjecting with additional outrageous jabs.

My advice to the original poster, was that it is OK to have overscan and that spec compliant video formats typically produce the best image quality. You disagreed with me, then kept insisting that modifying the output to get rid of overscan still produces spec compliant 720p, then started going on about some artificial 1152x648 to 1280x720 upscaling that will loose image quality if you don't underscan the output, and on and on - yet now you are saying I need to be more open. Then to top it off you still insist that I don't know my "own facts". Please - grow up a little.

Here's an interesting question for ya... are most movies these days really shot in Blu-Ray format or 720 x 480 and then upsized even though pixels cant be added out of nowhere that don't even exist as far as details and all goes?

karrih

06-16-08, 02:35 AM

Here's an interesting question for ya... are most movies these days really shot in Blu-Ray format or 720 x 480 and then upsized even though pixels cant be added out of nowhere that don't even exist as far as details and all goes?

I guess this is off topic, but most movies are shot on film. In order to make Bluray or DVD discs film is scanned with higher resolution than Bluray is. If modern digital camera is used, it also has at least the same resolution as Bluray.

madshi

06-16-08, 03:34 AM

With modern displays, I'm confident that there is not a significant number of pixels actually hidden by the bezel; it is more of a situation where they are "virtually hidden by the bezel". My explanation is based on a display that does not have 1280x720 available "pixels", but instead has a slightly smaller subset of pixels (this is typical of a the 720p displays I've worked with)

I honestly don't know if the very latest 1080p and 720p displays have fewer than 1920x1080 and 1280x720 physical pixels, but older ones certainly did. With my HDTV, when 720p is used, only 704 horizontal lines are visible for example. This is still a 1:1 mapping, but everything beyond the center 704 is not visible.
That explains why you and jong1 don't understand each other.

Usually when the term "overscan" is used with a fixed pixel display, zooming is involved. However, if your display actually has less pixels than 1280x720 and uses the missing pixels as the means to apply overscan then you're right that you can still have 1:1 pixel mapping even though the display is doing overscan. This sounds like a very odd display technique to me, though.

To my best knowledge most modern HDTV displays do not apply overscan like that. They usually zoom instead. They have to because they usually have a true physical resolution of 1920x1080. This is especially (but not only) true for front projectors, where LCD panels or DLP chips come with full 720p or 1080p resolution. The latest displays/projectors have an option to disable zooming/overscan. That is then called e.g. "dot-by-dot" mode. Blu-Rays are usually fit for displaying every pixel. So overscan is not needed for Blu-Rays.

For someone who previously insisted that mapping 0-255 into 16-235 and back to 16-235 unacceptably looses color accuracy
You mean mapping 16-235 to 0-255 and back to 16-235, I guess? That process doesn't necessarily lose color accuracy, if done in a perfect way. You do lose BTB and WTW, though. And as we all know, often things are not done in a perfect way. So the general rule to any video processing is that any processing should only be applied if really necessary. Mapping 16-235 to 0-255 and back to 16-235 is definitely not necessary. It's not helpful, either. So it should be avoided.

There is a reason why SDI modded DVD players are still widely considered to offer the best DVD image quality (when connected to a good video processor).

leeperry

06-16-08, 05:08 AM

Is that publically available? :-)

so that's the original BIOS, with random black screens in windows :

http://thumbnails.keepmyfile.com/46/1213601262fdc5db.gif (http://www.keepmyfile.com/image/fdc5db2257628)

that's the update I found on a chinese board(known issue there) :

http://translate.google.com/translate?u=http%3A%2F%2Fwe.pcinlife.com%2Fthread-802453-1-26.html&sl=zh-CN&tl=en&hl=en&ie=UTF-8

http://thumbnails.keepmyfile.com/46/121360126255041a.gif (http://www.keepmyfile.com/image/55041a2257627)

and that's the update eutech---sapphiretech---com sent me(newer ATi BIOS version, much more stable in video apps such as Haali's Renderer ) :

http://thumbnails.keepmyfile.com/46/1213601261f0be4b.gif (http://www.keepmyfile.com/image/f0be4b2257626)

only one last issue, sometimes my mouse pointer is corrupted badly.......but it goes away after a few secs(the hardware pointer acceleration seems fishy)...with older BIOS it would stay a lot longer :rolleyes:

but even with all those problems(the tasktray icon of drivers >7.11 also dies for no reason after 1/2 days), this HD2600 looks so much sharper than the nvidia counterparts on my CRT and my pj :eek:

but these ATi cards have a strong "prototype" feeling to them...talk about beta-testing :D

Dee_NA

06-16-08, 06:21 AM

Not sure if this off the subject or not but since im usin a DX9.0c for the HD 3650 on the XP, I keep hearin reports that there's gonna or already is DirectX10 for XP out. Anyone know if this is true and if this is done by some other programmer?

arfster

06-16-08, 06:58 AM

You mean mapping 16-235 to 0-255 and back to 16-235, I guess? That process doesn't necessarily lose color accuracy, if done in a perfect way. You do lose BTB and WTW, though.

If you do it via the CCC video controls at 16/86, you're actually contracting before rgb conversion (which is where the expansion is), so no btb/wtw loss. Doing it via the desktop colour controls, and I'm pretty sure the dongle, does it the other way though round though, so btb/wtw are lost.

It's visually lossless via CCC video controls though, from my tests (posted some screenies a while back, comparing some Bluray with ffdshow RGB output).

madshi

06-16-08, 07:40 AM

If you do it via the CCC video controls at 16/86, you're actually contracting before rgb conversion (which is where the expansion is), so no btb/wtw loss. Doing it via the desktop colour controls, and I'm pretty sure the dongle, does it the other way though round though, so btb/wtw are lost.
Note sure where exactly in the processing chain the CCC video controls are located. If it's before DXVA it's still in 8bit. In that case contracting the video data must come at a quality loss. Maybe it's not easily visible, but mathematically there must be a loss if you contract in 8bit. If it's done in 10bit then it probably only works with hardware accelerated video decoding. Now I'm personally using software decoding for MPEG2 and VC-1. And the CCC video controls never seemed to have any effect on the renderer and media player combination I'm personally using.

I stand by what I said: *Any* conversion should be avoided, unless it's really necessary. That's what every experienced ISF calibrator will tell you. I've been in the video processing forums for years. And the top rule for image quality has always been to avoid processing where possible.

arfster

06-16-08, 07:57 AM

I stand by what I said: *Any* conversion should be avoided, unless it's really necessary. That's what every experienced ISF calibrator will tell you. I've been in the video processing forums for years. And the top rule for image quality has always been to avoid processing where possible.

Sure, sure - I've been saying that here for a year now :-)

Not a tricky conversion though, CCC video controls just fiddle the luma after all. RGB conversion then just maps to a different set of values. Bit pointless all told, especially when it would be ridiculously easy to have an option for a "normal" levels RGB conversion matrix (after all, that's what usebt601csc is). I've been arguing this point with ATI tech support since Jun07 now......

However, in practice, it's not noticeable. Screenshots of the test available here:
http://www.avsforum.com/avs-vb/showthread.php?p=13914824#post13914824

As you can see, definitely not mathematically lossless, but you can't see the difference (and that's even when flicking between screenshots, and knowing where exactly in the picture to look).

madshi

06-16-08, 08:03 AM

As you can see, definitely not mathematically lossless, but you can't see the difference
Well, but it doesn't work for my media player / renderer / decoder combination (MPC HC, VMR9 Fullscreen, DScaler IVTC / MS VC-1). The CCC controls just don't have any effect in my playback chain.

leeperry

06-16-08, 11:32 AM

is there a simple way to output 16-235 on the DVI output ?

you got me lost with your endless discussions :D

I'm starting to realize that outputting 0-255 with 16-235 material creates banding :(

you can use a "Deband" filter, but well...the best option seems to output 16-235 from the PC(ie. ffdshow, windows desktop etc...)

so the only way to is to mess with the levels in CCC ?
is it done in 30 bit ? :D

arfster

06-16-08, 11:56 AM

is there a simple way to output 16-235 on the DVI output ?

Set CCC video controls to brightness16/contrast86, usebt601csc=1.

Or use no hardware features at all, and do RGB conversion in ffdshow (set ffdshow to output RGB32HQ, set to "full range"). That means no expansion in the first place, which is a lot simpler :-)

leeperry

06-16-08, 12:04 PM

well, at this point ffdshow does sharpening(using the multithreaded LSF avisynth script) then TV>PC conversion in RGB32HQ to HR.

but I have to use a deband filter prior to PC>TV conversion to avoid banding(220 values mapped on 256 simply doesn't work...).

so I could set ffdshow to output 16-235, but the windows desktop will still be in 0-255 :rolleyes:

arfster

06-16-08, 12:09 PM

Use the CCC desktop colour settings to contract back to video levels, that affects everything. It's not 16/86 though, uses a different scale. Piece of cake to calibrate though, use a 0-255 jpeg.

leeperry

06-16-08, 12:31 PM

well the CCC "color" settings also mess around with HR.

ideallly I'd like to output 16-235 to HR w/o any conversion whatsoever, so no need to touch anything for that.

and make the windows desktop 16-235 as well......but that doesn't seem possible.

jong1

06-16-08, 01:55 PM

That isn't possible to my knowledge (and we discussed it in great, some might say a sad amount of depth). But the ATI HDMI dongle, which also forces 16-235 and outputs YCbCr might get round your HR problem. It avoids any tinkering in CCC, but I'd probably only advise that if you have an HDMI display. Could always give it a go with a DVI HDTV though.

HT Slider

06-16-08, 02:34 PM

and make the windows desktop 16-235 as well......but that doesn't seem possible.

As Arfster mentioned this can be done using the "color" brightness/contrast settings (not AVIVO). With my DVI HDTV, to get just about everything outputting 16-235 (including the desktop):

Set CCC "color" brightness/contrast to +31/73%, add UseBT601CSC=1 and for most software leave the AVIVO brightness/contrast set to default.

The only software I have that ignores this is PowerDVD Ultra (running in Vista). To make PDVD Ultra output 16-235, change the PDVD advanced video brightness/contrast to +19/-5 (when playing an HD-DVD or Blu-ray).

One final oddity is I found I needed to use an ATI HDMI dongle and an HDMI to DVI cable, instead of simply using a DVI to DVI cable.

HT Slider

06-16-08, 03:55 PM

For anyone contemplating replacing their 2X00 series card with a 3X00 series card...

Last night I replaced my ATI 2600XT with an HIS 3870 IceQ3. Before doing so I rand through a number of different calibration samples, found and queued up several good and bad de-interlacing examples.

Comparing the 2600XT to the 3870, the video processing seems to be 100% identical. Anything the 2600XT struggled to deinterlace properly, the 3870 did also; anything it did well, the 3870 also did well.

Using the HDMI dongle (correct one for each card), the grey level performance appears to be 100% identical too.

Fan sound levels are if anything quieter with the 3870 IceQ3, possibly because the processor fan is able to slow down since the video card exhausts directly from the case.

Gaming performance on the other hand is absolutely incredible with the 3870. Grid and Crysis were borderline unplayable with the 2600XT. With the 3870 I can crank up everything and run much higher resolutions and frame rates are always greater than my HDTV. I really haven't tried all that much gaming wise yet, but this definitely brings our HTPC much closer to an all around entertainment system.

One thing that is odd is the 3870 is recognized by Catalyst as an RV630 chip, the same as the 2600XT. Searching the web, this seems to be normal.

I wonder if the video decoding/deinterlacing performance with the 4850 will be better or the same as the 2600XT?

digitlman

06-16-08, 04:09 PM

I just wanna make 100% sure that the 2400 pcie card can accelerate all bluray movies at all bitrates 100% perfect. i have had a problem setting this thing up for a friend (my 2600 agp card works perfect for me) i couldnt get accell to work with 8.5 drivers, i went back to the original 7.7 drivers that came with the card to get accell to work, its working, with low cpu usage but everything is jittery like it cant keep up.
This is with 1920x1080 output on the secondary display.
I think i am going to try the 8.5 drivers again as i think i understand the proper way to get the reg fixes in to hopefully make the accelleration work. (i used the 001 dir the first time, not the 001/umd/dxva)

thanks

leeperry

06-16-08, 04:22 PM

I'm using a DVI/HDMI Monster cable on a HC3100.

I will try in a few mins to send 16-235 HR to the pj(no conversions whatsoever, just plain 16-235 from ffdshow).

that will work fine I think, yet Windows will still be in 0-255 :(

I'm well aware that I could convert my video from TV>PC, then "undo" it with CCC or the projector OSD.....but my goal is to achieve no conversions, just plain 16-235 vanilla all the way :D

arfster

06-16-08, 05:07 PM

I'm well aware that I could convert my video from TV>PC, then "undo" it with CCC or the projector OSD.....but my goal is to achieve no conversions, just plain 16-235 vanilla all the way :D

The problem is that any driver-based desktop levels fiddling happens in RGB, and thus will always affect video as well.

What you're looking for would be an option in Windows itself, to actually pre-contract everything before its even fed to the mixer. I'm surprised this doesn't exist in MCE already, to be honest.

jong1

06-16-08, 05:28 PM

I'm well aware that I could convert my video from TV>PC, then "undo" it with CCC or the projector OSD.....but my goal is to achieve no conversions, just plain 16-235 vanilla all the way :DWe agree 100%!

This thread covered the same ground a little while ago. Although entitled "Nvidia" it talked a lot about ATI.

http://www.avsforum.com/avs-vb/showthread.php?t=1021811

leeperry

06-16-08, 05:47 PM

oh well I've started getting crazy over banding once I've decided to use the LimitedSharpenFaster avisynth script in ffdshow.

it looks *REALLY* good, but it lets you see right through ugly sources....and banding from TV>PC conversion looks obvious :eek:

anyhow, I've run a quick comparison between 16/235 and 0/255 and I'd say ffdshow does a much better levels conversion than my projector :D

I think the levels from the PC were fine ?!
RGB32HQ TV>PC conversion to HR, then the pj set to 0-255

and for TV level, I've disabled any levels conversion....so TV all the way, w/ the pj on 16-235

I've tried w/ and w/o "deband 1.2" in ffdshow.

the picture looks a lot more eye popping in PC level with deband(pj is D65 calibrated with a commercial sensor) :D

in TV levels, the deband filter looks too destructive(even lowering it) and the picture looks less contrasty.

hardly any pj can offer ffdshow-like performance..

here's a quote from some very knowledgeable x264 encoder :

1) Avisynth's crop() does not change image levels, sure as anything. That's a plain impossibility.

2) If you convert from TV to PC range in your AS script, do it after LimitedSharpen (or any other sharpening operation), never before.

3) TV->PC always amplifies present banding, and even introduces some new banding by itself. A method to make a more smooth conversion was posted here: http://forum.doom9.org/showthread.php?p=709492#post709492

4) With HD resolutions, the default operation of LimitedSharpen (with supersampling) has its issues. Rather use it without supersampling (i.e. ss_x=1.0,ss_y=1.0). Also, *do use* the soft=xx option. 40 to 80 should do, you have to try. Adjust overall strength as needed.

5) If there is some "motion" in that banding, Soothe() will be beneficial, too.

Regarding 3) -
- there's also an external tweak() plugin by prunedtree, which is said to apply dithering during changing levels. It's not documented anywhere, though.
- another possibility is to call gradfun2db before/or/after adjusting the levels. Danger is that the encoder will eat up the effect again.

here's what LSF looks like with deband and stuff in RGB32HQ, actually the deband filter increases the native contrast :eek: :

http://img241.imageshack.us/img241/7905/plople3.png