Thanks to all for their help.
Well I just got my R7 260x cards in today. Installed them both on each of my htpcs. I bumped up everything to madVR lvl5 and it looks good. It also helps to have a much quieter fan on the gpu now. All I can hear now is the fan from the projector which is right above my head.
Any suggestion as to what settings should be enabled in CCC? At the moment I think I just have pulldown detection and deinterlacing set. Is there anything else?
I have R9 295X2 graphics card and a 4K Panasonic TV.
What would be the optimal MadVR settings for the best possible picture quality?
Most content I have would be the following 4K upscaling:
1080p24->2160p24 (via HDMI, since DisplayPort only seems to support 2160p50 and 2160p60)
1080i25->2160p50 (via DisplayPort)
My R9 295X2 is primarily for gaming at 4K@60Hz, video upscaling is a bonus - so if MadVR sees only half of it then it would probably equate to whatever a single R9 290X can do.
Exactly the same OS (Win 8.1-64), same programs (MPCHC, LAV, reclock) and settings (Jinc, Jinc) on both. Works perfect on 750Ti, but starts to stutter once every second in exclusive mode after exactly 3 minutes (180 seconds) on 295X2.
- 4400 x 2250 x 60Hz = 594MHz
where 4400 = 3840 + 560 (Hblank), 2250 = 2160 + 90 (Vblank). As YCbCr 4:2:0 requires 12bit per pixel (= 8bit x 4 in Y+ 8bit in Cr +8bit in Cb / 4 pixels), TMDS bit rate is
- 4400 x 2250 x 60Hz in YCbCr 4:2:0 = 594MHz x 12bit/pixel x 10/8 (overhead) = 8.91Gbps
(This is the same bit rate as 4K 30Hz = 297MHz RGB 8bit.) HDMI transmitter in the current graphics cards supports
- 300MHz (pixel clock rate)
- 300MHz x 24bit/pixel x 10/8 = 9.0Gbps (TMDS bit rate)
that's enough for 4K x 2K x 60Hz in YCbCr 4:2:0. That's not enough for 4K x 2K x 60Hz RGB 8bit:
- 4400 x 2250 x 60Hz in RGB 8bit = 594MHz x 24bit/pixel x 10/8 = 17.82Gbps
Last edited by renethx; 06-21-2014 at 11:26 PM.
And another question. Many users are trying more than 32 neurons with NNEDI3, does it make sense with 1080p to 2160p? ( do not care about SD to FHD )?
- 4400 x 2250 x 30Hz = 297MHz.
This is half of 4K @60Hz, 594MHz. TMDS bit rate is
- 4K @30Hz RGB/YCbCr 4:4:4: 297MHz x 24bit x 1.25 overhead = 8.91Gbps
- 4K @60Hz YCbCr 4:2:0: 594MHz x 12bit x 1.25 overhead = 8.91Gbps
Both are within HDMI 1.4 specifications (hardware-wise).
1080p@24Hz -> 2160p by NNEDI3 32 is possible with R9 280X. 1080p@60Hz -> 2160p by NNEDI3 32 is possible with R9 290X.
Last edited by renethx; 06-22-2014 at 03:28 AM.
R9 270X can handle NNEDI3 32 1080p24 -> 2160p, but can't handle NNEDI3 (whatever neurons) 1080i/p60 -> 2160p.
R9 280X can handle NNEDI3 16 1080i/p60 -> 2160p.
R9 290X can handle NNEDI3 32 1080i/p60 -> 2160p.
There are some cases where 64 is better. You'd better try youself.
Last edited by renethx; 06-23-2014 at 12:55 AM.
According to this, and I have also verified this the settings running the "Highest" preset resolves in;
Chroma upscaling; Jinc, 3 taps (AR)
Image upscalnig; Jinc, 3 taps (AR)
Image downscaling; Catmull-Rom (AR, LL)
I'm not entirely sure whether these are considered the optimum settings any more? I've noticed the new NNEDI3 option which there before. I clearly got no clue about the difference between these options, but as I'm running a GTX Titan I can't see any reason why I should not select the best settings.
I'm running 1920x1080 @ 120Hz, what scaling and various other options in this jungle of MadVR settings are considered to be the optimum choice these days?
I have also noticed that my GTX Titan for some awkward reason is having a very hard time running MPC-HC @ MadVR + SVP (Smooth Video Playback). I know for a fact that SVP is supposed to be VERY intensive on the hardware, but running Intel Core i7-3960X + GTX Titan one would think my system should be able to handle both on maximum settings?
Might it be because of running re-clock + SVP is a rather stupid idea as re-clock tries to set the refresh rate of my monitor to match the one of the video (typically 23.976 Hz) at the same time as SVP is doing it's interpolation magic trying to interpolate the frames into 120Hz at the same time? I have no clue what's going on but with SVP enabled with the default setting it chooses will get slow downs for time to time. I have also noticed that SVP got a wide arrange of various settings as well and was wondering if anyone got some tips to what settings is considered optimum?
(These levels are just a starting point. You can configure madVR in various ways.) Titan should be able to handle SVP Level 5g + madVR Level 5 for up to 1080p24 in a 1080p 60Hz display without problem. Your display is 120Hz so YMMV.
Nowadays ReClock's only main purpose is send bit-perfect LPCM audio to the audio device because GPU supports very precise '23.976Hz'. Then I feel why not use MPC-BE? (Of course you can use ReClock for several other purposes.)
in my case:
asus essence STX
i normally have a 23p clock of ~23.97663 so an ok clock but i get a drop every ~25 min
a drop means the clock is to slow it should be 23.976023976... but 23.97653 is to slow?
with a custom resolution resulting in 23.7736 i get a repeated frame every 7-10 h.
the problem with these tweaks is i still haven't found a way to to get a custom 23p and a custom 24p resolution... the 23.976(23.7736) is treated as 24 hz
Last edited by renethx; 06-24-2014 at 04:28 AM.
I've tried almost every media center incl. MediaPortal, XBMC, Plex and MediaBrowser. So far MB is the most promising, but MBT is rather limited in the early stage. MediaPortal 2 is also in beta and the development is really really slow, there is not even native iOS remote yet. Both MP and MBT use LAVF, MBT even allows MadVR which is great for PQ. XBMC is small, powerful and very stable with minimal configuration, but it is based on thick client with no centralized server and no support for streaming. So far, none of the mediacenter can replace PowerDVD and TMT for Bluray playback.
I'm interested to know what is your setup?
if you have one with multiply hour by default you are just very lucky X.-)
I would imagine almost anyone could find happiness with one of those options^. And you can even use them all, in different places. That is the strength of MB server IMO. I could have one of each type of system in my house, or I could have an HTPC that had all three solutions on it and they would work seamlessly. No reason you need to only have 1. Without MB3 server you would lose a lot of functionality and multi product support for tablets, ROKU, phones and all that too. You'd give up transcoding and remote access too. MB3 project is advanced past PLEX at this point.
The ROKU app craps all over the PLEX ROKU app. One last thing left is for MB3 to fix the Fast Forward and Rewind (in process) and after that it's really not a contest which is better. PLEX developers blame the limited ROKU OS code but the MB3 ROKU app in the beta channel proves a lot more can be done with good developers and creative out of the box thinking. I could write a book on the small details the ROKU app for MB3 gets right while the PLEX app gets it wrong. MB3 server does so much more in terms of metadata, notifications, trailers, actors images and bio, metadata management, it has a way nicer user interface too. MB server is clearly better than PLEX server. And I think MB front ends are better than the PLEX front ends generally. At first I used PLEX more because it was straight forward and just works, but recently MB is emerging as the clear favorite - it is also becoming a lot more straight forward and easy to use and set up- and the beta problems are fixed quickly and the project keeps advancing rapidly. XBMC or PLEX never showed the promise that MB3 does, and the lightning fast blitzkrieg of apps and features is amazing and doesn't seem to stop. Even long term hardcore XBMC front end users should find value in MB3 platform.