or Connect
AVS › AVS Forum › Video Components › Home Theater Computers › Which HD **** would be sufficient for MadVR
New Posts  All Forums:Forum Nav:

Which HD **** would be sufficient for MadVR - Page 5

post #121 of 207
Oh.
post #122 of 207
Quote:
Originally Posted by StinDaWg View Post

I see no difference between MadVR and XBMC using Lanczos3 or Spline 36 scaling.

Precisely speaking, there is a difference: anti-ringing filter is implemented only in madVR. StinDaWg just can't see the difference between LZ+AR and LZ (neither can I in real 720p movies).

MadVR has a more sophisticated algorithm, Jinc.

- LZ: sharper, coarser
- Jinc: softer, finer

The difference is evident, but some people (like StinDaWg) prefer the sharper algorithm. Personally I prefer Jinc for all contents (SD, HD in FHD screen) because it looks a lot more natural.
Edited by renethx - 8/19/13 at 12:40am
post #123 of 207
Jinc also has less aliasing artifacts:



(Monsters Inc. Blu-Ray Upscale to UHD)
post #124 of 207
I need to get a more powerful card than the GT-430 I have now. What is the current recommendation?

I also applied madshi's 0-256 fix to my HTPC a while ago. If I go with AMD, how do I undo that and do I even need to undo that?
post #125 of 207
I don't think you'd need mess with it with a Radeon card

It's a totally different card, and driver.

Just uninstall the old one and you are good.
I would start here. Then test it quickly and only mess with it if you find trouble.

As far as cards- what specifically do you want to do ? Only MadVr?

I think a 7000 series Radeon for $79-129$ is a good value proposition
post #126 of 207
Quote:
Originally Posted by Mfusick View Post

I don't think you'd need mess with it with a Radeon card

Right, but I already did mess with it for my nVidia card using the utility that madshi has to do this. It is a bat or exe that modifies the registry for you so you don't have to hack it yourself. If I go from nVidia to AMD I will need to undo that hack.
post #127 of 207
This post should be helpful.

madshi's program has "undo" option too. Perhaps it won't affect AMD card, however.
post #128 of 207
Quote:
Originally Posted by Sammy2 View Post

Right, but I already did mess with it for my nVidia card using the utility that madshi has to do this. It is a bat or exe that modifies the registry for you so you don't have to hack it yourself. If I go from nVidia to AMD I will need to undo that hack.

You are quick biggrin.gif

I added to my previous above but I think you are ok. Undoing a registry hack should not be needed but doing so isn't that hard either.

I think it's best to just try the card first and see what you get before you go messing around in registry.
post #129 of 207
Quote:
Originally Posted by Mfusick View Post

I don't think you'd need mess with it with a Radeon card

It's a totally different card, and driver.

Just uninstall the old one and you are good.
I would start here. Then test it quickly and only mess with it if you find trouble.

As far as cards- what specifically do you want to do ? Only MadVr?

I think a 7000 series Radeon for $79-129$ is a good value proposition

It is an HTPC only. No gaming. I do watch Live/Recorded CableTV with it too.
post #130 of 207
Quote:
Originally Posted by renethx View Post

This post should be helpful.

madshi's program has "undo" option too. Perhaps it won't affect AMD card, however.

That is a great post! I think I saw that before. I bookmarked it this time.

Do you know the name of the program so I can find it in my d/l folder on my HTPC when I go to undo the settings?
post #131 of 207
Quote:
Originally Posted by madshi View Post

Jinc also has less aliasing artifacts:



(Monsters Inc. Blu-Ray Upscale to UHD)

The nearest does seem to have a noticeable amount of pixelization, but for the rest, I don't see any difference that makes it worth the while to shell out $$$ for a discrete highend GPU and spend time tweaking... just my opinion BTW...
Once again, I didn't say there is no difference... I just couldn't care less about it...
post #132 of 207
Quote:
Originally Posted by Sammy2 View Post

Do you know the name of the program so I can find it in my d/l folder on my HTPC when I go to undo the settings?

madNvLevelsTweaker.exe, included in madVR (works outside madVR too). Check / uncheck "Force PC levels (0-255)" and restart your system.
post #133 of 207
x4 enlargement (this forum resizes every picture posted). This typically shows how Jinc is good.


Edited by renethx - 8/19/13 at 6:47am
post #134 of 207
Quote:
Originally Posted by balky View Post

I don't see any difference that makes it worth the while to shell out $$$ for a discrete highend GPU and spend time tweaking

Look at this table. HD 6570 is enough to get the highest quality setting of madVR if you are interested in upscaling movies in a FHD screen. To upscale older SD music video and documentaries, you will need HD 7770, however (it's a mid-range graphics < $150).
post #135 of 207
Quote:
Originally Posted by Sammy2 View Post

I also applied madshi's 0-256 fix to my HTPC a while ago. If I go with AMD, how do I undo that and do I even need to undo that?

That fix is NVidia specific and doesn't affect any other GPUs.

Quote:
Originally Posted by balky View Post

The nearest does seem to have a noticeable amount of pixelization, but for the rest, I don't see any difference that makes it worth the while to shell out $$$ for a discrete highend GPU and spend time tweaking... just my opinion BTW...
Once again, I didn't say there is no difference... I just couldn't care less about it...

That's just fine. Everybody needs to decide for himself.
post #136 of 207
Thanks!!
post #137 of 207
Quote:
Originally Posted by StinDaWg View Post

Just my opinion but I completely disagree. I watch a lot of 720p tv and Blu-ray rips. I see no difference between MadVR and XBMC using Lanczos3 or Spline 36 scaling.
Lanczos and Spline produce very similar results. Jinc AR is the next step up from Lanczos/Spline in image quality. And personally I think Lanczos/Spline are too sharp for lower resolution video, and make compression artefacts too obvious.
Quote:
Originally Posted by madshi View Post

Jinc also has less aliasing artifacts:
http://www.avsforum.com/content/type/61/id/260659/width/800/height/600
(Monsters Inc. Blu-Ray Upscale to UHD)
Good example - what are your chroma settings? It looks like it might be a little soft. (may just be the source?)
post #138 of 207
Although it is soft, Jinc3 does not have the jaggies apparent in Lanczos3.
post #139 of 207
Quote:
Originally Posted by renethx View Post

x4 enlargement (this forum resizes every picture posted). This typically shows how Jinc is good.

This is a perfect example of why I prefer Lanczos/Spline over Jinc. Jinc is a lot softer and you can't make out the individual hairs as easily, they just blend together. It's a bit blurry. Some may prefer that but I don't. In real time playback the difference is not as noticeable (I don't see any ringing or artifacts in 720p->1080p scaling) and I don't have a 4K display. We're all just splitting hairs here anyways (pun intended). biggrin.gif
post #140 of 207
Quote:
Originally Posted by StinDaWg View Post

This is a perfect example of why I prefer Lanczos/Spline over Jinc. Jinc is a lot softer and you can't make out the individual hairs as easily, they just blend together. It's a bit blurry. Some may prefer that but I don't. In real time playback the difference is not as noticeable (I don't see any ringing or artifacts in 720p->1080p scaling) and I don't have a 4K display. We're all just splitting hairs here anyways (pun intended). biggrin.gif
Well you are looking at an upscaled image, of an upscaled image, and it seems like it might have originally been scaled using a soft chroma algorithm.
Here's what madshi's original image looks like upscaled to 8x. (the "nearest" image was downscaled 50%, then upscaled 8x using each algorithm)
y42r9uko.jpg


Jinc does much better than most of the other algorithms in madvr for aliasing, and that's a big deal, especially when upscaling to 4K or beyond.
Spline/Lanczos are too sharp and exaggerate any ringing or compression in the source. Both suffer from aliasing at the lower values. Jinc is relatively neutral in that regard.
Jinc 4 may be preferred over Jinc 3 at these large scale factors though, as it should be a little sharper - though it does exhibit more ringing.
post #141 of 207
Quote:
Originally Posted by Chronoptimist View Post

Spline/Lanczos are too sharp


I prefer a sharper image for 720p->1080p upscaling, while you prefer a softer, more natural image. I'm willing to have some minor ringing that I don't see in normal viewing to get that. Neither of us are wrong.
post #142 of 207
If you have to upscale to 8x to see a difference on computer generated hairs against a dark background then it might not be worth it (added $$$, heat, space, noise, effort, frustration, e-waste, etc) to all.

Been saying it for years now. Don't expect everyone to see a difference, please don't tout it as some amazing advancement for *everyone* and that if you don't see a vast improvement then you "don't know what you are talking about". With that being said I understand many do see a difference --- whether its always real or sometimes placebo is another debate.

Just my $.02. I'll get off my soapbox now.
post #143 of 207
If you feel so strongly, then why on earth would you write a guide.
post #144 of 207
Quote:
Originally Posted by Andy_Steb View Post

If you feel so strongly, then why on earth would you write a guide.
I write guides on all sorts of things when I am testing them and trying them out.

Obviously you didn't read the disclaimer at the beginning of my madvr guide. wink.gif
post #145 of 207
Quote:
Originally Posted by assassin View Post

If you have to upscale to 8x to see a difference on computer generated hairs against a dark background then it might not be worth it (added $$$, heat, space, noise, effort, frustration, e-waste, etc) to all.
The difference is clear to me in the 2x example, and SD to 1080p is roughly a 2.5x scale. In fact I have some videos which are 3x. SD to 4K is 5x. Chroma is always double that, so scaling factors of 8x are actually relevant.
Quote:
Originally Posted by assassin View Post

With that being said I understand many do see a difference --- whether its always real or sometimes placebo is another debate.
There is nothing to debate - there are demonstrable benefits to madvr scaling.
If you want to argue that your TV is too small or that you're sitting too far back to see the differences... well, that's your loss.
post #146 of 207
@Chronoptimist, there's no use discussing this with assassin. He's made up his mind a long time ago that madVR would be mostly placebo and no amount of good reasoning is going to change his opinion, even if you throw in some hard evidence like screenshots. Just some examples:

(1) He claims that you need an 8x factor to see a difference, while the difference is already clearly visible in the 2x screenshots.
(2) Seemingly he doesn't understand that watching screenshots on a tiny computer monitor is different to using e.g. front projection.
(3) He still claims you need an expensive dedicated and hot GPU to use madVR, while he was told many times that madVR runs just fine on newer integrated GPUs (with some mild compromises).
(4) Seemingly he isn't aware of that the display market is going to 4K sooner or later (similar to the Megapixel race with digicams), which means 2x scaling for Blu-Ray and much higher scaling factors for SD.
(5) He doesn't even acknowledge that there *are* factual benefits to madVR's scaling algorithms, he just grateously "allows" that some people might see a difference, but not without throwing in a hint that it might be placebo.
(6) He doesn't acknowledge that there are more benefits to madVR than just scaling.
(7) He says "Been saying it for years now", ignoring the fact that madVR got many new algorithms in the last ~ 1.5 years (IVTC, anti-ringing, Jinc, Smooth Motion FRC, ...) which are technically clearly superior to what any other renderer does.
Edited by madshi - 8/20/13 at 12:27am
post #147 of 207
I never said there was *no* difference. I said the difference is small to some and might not be worth the trouble to *all*. Statements saying there are demonstrable benefits - period - are just wrong. There are demonstrable benefits to *some* and some prefer other algorithms altogether (as evidenced in this thread and others).

Since when were we discussing 4k? Obviously that might be a whole different game.

You are correct in saying that often times it's free (from a hardware standpoint) to try. However, again, this is not the same for *everyone* and certainly not the case for the better algorithms you are touting.

Ps: I have a 120"'screen.
post #148 of 207
Quote:
Originally Posted by assassin View Post

You are correct in saying that often times it's free (from a hardware standpoint) to try. However, again, this is not the same for *everyone* and certainly not the case for the better algorithms you are touting.

I wasn't talking about "everyone". My point is that you're always complaining about the heat/power/cost needed to use madVR. Which doesn't make a lot of sense because you don't need a dedicated GPU to benefit from madVR, as you have been told a dozen times already. For example, I've just played an NTSC DVD with the following madVR configuration:

(1) IVTC with decimation (which DXVA cannot do!).
(2) Upscale to my computer monitor's native resolution (1680x1050) with Jinc 3-taps with anti-ringing filter.
(3) Smooth Motion FRC enabled to get rid of the 3:2 pulldown judder (my display only supports 60Hz).

Worked just fine with zero frame drops and perfect smoothness. Note that all of this was done with an HD4000 integrated Intel GPU. And I was using 3 (!) high-quality algorithms which no other PC playback solution can currently compete with.

I've recently asked some of my guests whether they were able to see a difference in motion smoothness between Smooth Motion FRC on/off with my 60Hz computer monitor, with a 24fps movie. All of them could easily see the difference (with camera pans or similar scenes). So there you have your "demonstrable benefits".
post #149 of 207
I think you just proved my original point.
post #150 of 207
I truly believe there is a difference, which for some, out of curiosity might be worth trying to verify if you will not need to break the bank or live with remarks like "it was at least working ok until you thought you could make it better"

All my HTPCs are running Linux with XBMC and I rarely watch SD stuff...
There is plenty of SD stuff for the kids though... Teletubies, Tom & Jerry... etc... but believe me, they don't give a flip about image or sound quality...

For me, I don't see myself re-installing my HTPCs in windows just for this difference... for some it might be worth the effort...

On the other hand, I would like to try mediabrowser 3, (just waiting for the Linux version of server to be out)
The combination of the two factors is enough to make me switch back to windows on HTPC if the result of testing comes out as advertized. i.e. the mediabrowser 3 eye candy thingy and the whatever level of Jinc my GPU will handle...

I just think it is better to let each one decide for themselves what they want and if the effort is worth it or not... and without any pre-prejudice... tongue.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Home Theater Computers
AVS › AVS Forum › Video Components › Home Theater Computers › Which HD **** would be sufficient for MadVR