Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
13th November 2015, 11:27 | #34243 | Link |
Registered User
Join Date: Oct 2015
Posts: 5
|
I really can't spot an enormous difference while comparing "crispen edges" and "sharpen edges" with 1080p content on a 1080p display when it comes to sharpness. It looks like there's indeed a difference when it comes to brightness. "Crispen Edges" seems to brighten the image, "Sharpen Edges" does nothing of that kind and that's why I personally favor "Sharpen Edges". "Shapren Edges" produces clearly a more natural looking image to me. There's just an enormous drawback. The algorithm consums about 7ms of render time on my GTX 960 (not OC) and that's just too much in combination with my other settings I don't want to adjust.
It would be great if it is possible to optmize the algorithm so that is just needs about 3ms of render time, maybe 4ms. Hey, at least I'm free to dream. I have a question reagarding DXVA2-copyback decoding. Would there be a huge performance gain if I replace - say a DD3-1300 CL 9-9-9-24 RAM - with DDR3-1600 7-8-8-24 modules? Last edited by Murmeltier; 13th November 2015 at 11:43. Reason: Changed the word "brighteen" to "brighten" |
13th November 2015, 12:23 | #34246 | Link | |
Registered User
Join Date: Oct 2012
Posts: 7,925
|
Quote:
|
|
13th November 2015, 17:34 | #34247 | Link |
Registered User
Join Date: Dec 2011
Posts: 1,812
|
Some more thoughts on SuperRes radius:
I think one should simply use a lower SR strength for very high scaling factors, e.g. not above 2. With it, I think aliasing is not an issue with real content in the wild and with high scaling factor, the effect of SR subjectively looks stronger to me in general. So imho a radius of 0.66 is fine in general and if one wants less aliasing but still a nice sharpness boost, SR strength of 2 should be combined with the new sharpen options (at least once they are more affordable). I would find it really bad if the radius was higher, as I really like the naturally sharp impression when doing 1080p -> WQHD. I suppose a radius of 0.66 also looks fine with 1080p -> 4k, as the aliasing seems to depend on a very low pixel count of thin structures. |
14th November 2015, 01:44 | #34248 | Link | ||
Registered User
Join Date: Apr 2006
Posts: 299
|
Quote:
Quote:
|
||
14th November 2015, 19:28 | #34249 | Link | |
Registered User
Join Date: May 2015
Posts: 106
|
Quote:
There's another isssue with those interlaced DVDs since the introduction of the automatic switch to DXVA scaling: When playback starts the clock deviation is heavily off, maybe 10^6 % or so, and the DVD essentially hangs for 10 to 20 sec until this value goes down to an acceptable level and plays fluidly thereafter. |
|
15th November 2015, 14:52 | #34250 | Link |
Registered User
Join Date: Aug 2007
Posts: 79
|
hi there madshi. first time I'm posting here but I'm using your VR for a long time now. I was playing around with the new implemented sharpen agorithms and so far I'm pretty satisfied with sharpen and crispen, both at default values. the effect with superres is too subtle though for my taste, but maybe my display resolution is too low for the filter to be useful. I usually watch 720p content with a native screen resolution of 1680x1050. regarding luma and adaptive sharpen, I also don't see much use for them. I can't imagine content where I would prefer them over the new four ones.
anyways, I wanted to draw some attention to a new developed downscaling algorithm by ETH Zurich which is called "Perceptually Downscaling" which gets pretty amazing results: https://graphics.ethz.ch/~cengizo/imageDownscaling.htm I think that could be a gread addition to madvr. according to the paper the performance of is pretty good (they implemented it with matlab) and it should be possible to execute it in realtime. it seems that they didn't publish the code though but the algorithm is explained in the paper and presented in pseudo code, it only consists of convolutions and sums. Last edited by Sp00kyFox; 15th November 2015 at 15:22. |
15th November 2015, 16:23 | #34252 | Link |
Registered User
Join Date: Jun 2007
Posts: 656
|
I concur, if I have to limit myself at the images displayed on that page, the algorithm seems pretty terrible, to be honest. I like a sharp image, but judging from these images one is supposed to watch a Minecraft game, not a movie.
Edit: pictures here are better, though https://graphics.ethz.ch/~cengizo/Fi...ralImages.html Aliasing is still there, but at least I can see a reason for taking it into consideration. They really ought to change the pics in the homepage. Edit 2: these https://graphics.ethz.ch/~cengizo/Fi...udyImages.html are even better. Interesting... Last edited by ashlar42; 15th November 2015 at 16:27. |
15th November 2015, 17:03 | #34253 | Link |
Registered User
Join Date: Mar 2006
Posts: 567
|
Bug report: with the provided file (35Mb), madVR breaks in many different ways when using QuickSync as the decoder in MPC-HC. EVR (custom presenter) has no problems, with the same decoder. It's a straight mux to mkv of the first minute of the first episode of Star Trek Voyager, from a PAL DVD disc (25i material).
I am listing the steps to follow for what I describe below, because based on what you do, the issues you get are different: - connect your display (TV in my case) directly to your motherboard HDMI port; if you connect your display to another graphic card, and just use QS decoding in the background, the results below will be different (like, no issues in some cases and much worse issues in others); - switch your display to a 50Hz refresh rate (again, if you have a different refresh rate, the issues are much worse) - reset the madVR settings, using the "restore default settings.bat" in the madVR folder - in MPC-HC Options: Playback\Output\Auto-zoom = enabled, and set to AutoFit (Larger only) - in MPC-HC Options: Playback\Fullscreen\Launch files in Fullscreen = disabled Now, in MPC-HC, let's first configure to use EVR as the video renderer, select xySubfilter as the subtitle renderer, and QuickSync as the hardware decoder, and play the file. During playback, make sure to try and skip back and forth in the file (using the keyboard arrows, for instance), and to switch between windowed mode and full-screen mode, just to make sure that everything works fine. Notice that subtitles are located nicely at the bottom of the image, too. - with "Enable Adaptive HW deinterlacing = disabled" -> no issues - with "Enable Adaptive HW deinterlacing = enabled" -> no issues Now select madVR as renderer, and make sure xySubfilter is also selected as the subtitle renderer. Start playback, but pause it immediately to configure madVR a bit more: - devices->display modes->switch to matching display mode = disabled - processing->zoom control->move subtitles = disabled - rendering->general settings->full-screen exclusive mode = disabled. Playback results: - with "Enable Adaptive HW deinterlacing = disabled", madVR using DX9 video path, while the file plays in a MPC-HC window of the same size as the video material, everything seems to look fine; the opening credits are scrolling, and then the action in the video begins, with some talking heads; it's at this point that you see subtitles in the video, too; but try to switch to fullscreen mode -> the subtitles are no longer at the proper location in the video ! They're now in the middle-left part of the screen. - same as above, but with "Enable Adaptive HW deinterlacing = enabled": again, everything seems to be fine; let's try and switch to full-screen mode-> what do you know, this time the subtitles are in the proper location (at the bottom of the screen)! Don't forget to switch between full-screen mode and windowed mode, and to skip in the file back and forth, to make sure that everything else is fine! Ok, time to switch madVR to use the DX11 path: - rendering->general settings->use Direct3D 11 for presentation = enabled - rendering->general settings->present a frame for every VSync = enabled Now, playback results: - with "Enable Adaptive HW deinterlacing = disabled", madVR using DX11 video path, while the file plays in a MPC-HC window of the same size as the video material, everything seems to look fine, at least at the beginning; when the talking heads appear, you might notice some judder, but you might also think it's a fluke. So, after these talking heads appear, let's switch to fullscreen mode -> very likely you'll see an extreme case of frames jumping back and forth, making the chaotic video even more chaotic. Or, if per chance this does not happen, just skip back a bit, I guarantee the extreme jumping will start. And, did you notice the subtitles are once again located in the middle-left part of the image? - with "Enable Adaptive HW deinterlacing = enabled", madVR using DX11 video path, switching to full-screen does not have the same dramatic effect as above, you notice some jumps but not that bad. But it's enough to skip back or forth in the file, and the extreme frame jumping starts. I'll make one more comment, about having the display connected to another graphics adapter, instead of your motherboard video port. In this case, having "Enable Adaptive HW deinterlacing = disabled" might give a working picture, but turn this setting on, and this is by far the worst playback scenario - you might get a frame every 1 minute or so, if you have the patience to wait for it; we can't really say there is any playback in this case. Finally, switch back to EVR (custom presenter) as the video renderer, to have proof that everything can work fine when using the same decoder and the same file. These kind of issues don't happen with every file. But when it happens, it's really bad. Last edited by KoD; 15th November 2015 at 17:24. Reason: details |
15th November 2015, 19:34 | #34254 | Link | ||||||||||
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
Quote:
|
||||||||||
15th November 2015, 19:50 | #34255 | Link |
Registered Developer
Join Date: Sep 2006
Posts: 9,140
|
madVR v0.89.17 released
http://madshi.net/madVR.zip Code:
* significant speedups for "sharpen edges" and "thin edges" algos * small quality change (hopefully improvement?) for "thin edges" * fixed: repeated frames were reported although smooth motion FRC was on Code:
AMD HD7750 1080p24: AdaptiveSharpen: 6.34ms sharpen edges v0.89.16: 17.15ms sharpen edges v0.89.17: 9.87ms thin edges v0.89.16: 25.62ms thin edges v0.89.17: 5.41ms Intel HD4000 720p24: AdaptiveSharpen: 9.92ms sharpen edges v0.89.16: 36.2ms sharpen edges v0.89.17: 15.9ms NVidia GTX650 1080p24 AdaptiveSharpen: 15.51ms sharpen edges v0.89.16: 27.59ms sharpen edges v0.89.17: 11.69ms |
15th November 2015, 20:35 | #34256 | Link |
Registered User
Join Date: Dec 2011
Posts: 1,812
|
Thanks for the performance optimizations.
With GTX 980, now Adaptive Sharpen and sharpen edges roughly seem to be equally fast. Adaptive Sharpen is still slightly faster, I get a little lower GPU usage and rendering time with it. So there seems to be a difference to Kepler. Regarding SuperRes again: I read here that SR would produce an overly sharp picture to some when doing 1080p -> WQHD. However, I can hardly believe this. I made a sharpness comparison with a game screenshot rendered in WQHD, and despite of SR with strength of 3 on top of Jinc AR scaling, it's still noticeably softer than the original (and much closer to ground truth than just Jinc AR without SR): Original in WQHD: Downscaled to 1080p with C-R GL: Upscaled to WQHD Jinc AR without SR: SR strength 3: |
15th November 2015, 21:09 | #34257 | Link |
Kid for Today
Join Date: Aug 2004
Posts: 3,477
|
Well, 2X SR improves edges clarity so how about another extra pass on top of it for instance? Most users here don't care about increased GPU load if PQ improves.
Thanks a lot for .17 but apparently 2X is still disabled for SR only? Based on my screenshots comparison, you seemed to agree that 2X SR improves PQ so why not? I got no use for 1X only SR |
15th November 2015, 21:22 | #34259 | Link | |
Registered User
Join Date: Jan 2010
Posts: 169
|
Quote:
AFAIK the only cards that can do that are GTX 960/950. Maybe there are other platforms( integrated GPU/CPU) ,laptops that can do hevc hardware decoding. |
|
15th November 2015, 21:41 | #34260 | Link | |
Registered User
Join Date: Mar 2006
Posts: 567
|
Quote:
What I looked for was to have QS doing the decoding and deinterlacing, and having the GTX 980 doing the resizing. But decoding + deinterlacing on QS was not working properly when using the dGPU for output (as soon as I skipped in the file, playback started grinding), so I tried to debug what was going on, and connected the TV directly to the motherboard HDMI port. That's how I encountered all of the above. madshi, is it certain that the bat file really resets all the settings? And that it does not have to be run as administrator? Because I did not have the issue with subtitle positioning on this same file when I used 0.89.15 with the ar fixed dlls. But the issue started happening after switching to 0.89.16 (and doing a reset settings to defaults). Then I switched back to 0.89.15 + the ar fixed dlls, did a reset to defaults, and still had the subtitle issue. Finally, I started modifying all the settings in madVR, just enabling stuff, pressing Apply, and then disabling them back, and after some such random operations the subtitle issue was gone again with 0.89.15. But I also started doing the same in the Intel Graphics CP (like disabling the 3D Application Optimal Mode setting - when this is enabled the madVR OSD flickers badly on that file), so I'm not sure what fixed the positioning. The frame jumping back and forth is still there though, as are the issues when using the GTX 980 for output and QS for decoding and deinterlacing. Last edited by KoD; 15th November 2015 at 21:53. |
|
Tags |
direct compute, dithering, error diffusion, madvr, ngu, nnedi3, quality, renderer, scaling, uhd upscaling, upsampling |
|
|