madVR - high quality video renderer (GPU assisted)

[aufkrawall]

I think there's still an issue with NNEDI3 image quadrupling on Nvidia, occasionally driver still gets stuck in slow P2 state (but without any visible corruption).
Question is if it is caused by madVR or by the 352.xx driver series (Windows 8.1 x64), which seems a bit unstable.
I'd need to find a completely reproducible scenario to provoke the unwanted bahvior, I guess...

[6233638]

I think there's still an issue with NNEDI3 image quadrupling on Nvidia, occasionally driver still gets stuck in slow P2 state (but without any visible corruption).
Question is if it is caused by madVR or by the 352.xx driver series (Windows 8.1 x64), which seems a bit unstable.
I'd need to find a completely reproducible scenario to provoke the unwanted bahvior, I guess...I don't know that it's an issue with NNEDI3 specifically.
When doing some 4K testing, I was also running into an issue where my video card would stay in the medium power state (P8) rather than using the full power state (P0).

I think it managed to hit a certain GPU load which was not enough to trigger the higher performance state, but also too low to actually play videos without dropping frames.
Prior to 4K decoding you could use CUVID decoding to force the GPU into a full power state, but my GTX570 is too old to decode 4K so decoding is handled by the CPU instead.

I don't know what specifically fixed it, but changing the scaling settings to something a little more demanding was enough to push it into using the full power state all the time.
Presenting a frame every V-Sync may also have been a contributing factor. (but I can't get smooth playback with that option disabled no matter what format the video is)

[cyberscott]

Does this (32bit only) test build change anything?

http://madshi.net/madVR8810test.rar
I used MPC-HC 32 bit. No change. Here is the log from the test build.
http://s000.tinyupload.com/?file_id=81314808012845964113

[aufkrawall]

I meant P8 state, sorry.
madshi did something to fix NNEDI3 quadrupling in the past with image enhancement because there was a similar issue.
I suppose it's still related to NNEDI3 quadrupling + image enhancement.
Driver gets totally stuck in P8 state, thus it drops lots of frames if it's too slow. It's not able anymore to enter a higher P-state unless the driver is restarted.

[madshi]

I used MPC-HC 32 bit. No change. Here is the log from the test build.
http://s000.tinyupload.com/?file_id=81314808012845964113
How about this one?

http://madshi.net/madVR8810test2.rar

[huhn]

I meant P8 state, sorry.
madshi did something to fix NNEDI3 quadrupling in the past with image enhancement because there was a similar issue.
I suppose it's still related to NNEDI3 quadrupling + image enhancement.
Driver gets totally stuck in P8 state, thus it drops lots of frames if it's too slow. It's not able anymore to enter a higher P-state unless the driver is restarted.

can you rename the mpc-hc.exe? it's very unlikely it will fix the issue but maybe nvidia tries to use a low power profile for mpc-hc.

[aufkrawall]

The MPC HC profile has some entries, whatever they mean.
Gonna delete it. I accidently also had prerendered frames limited to 1. Dunno if this makes any difference, since I don't use FSE, but I'll give it a try.
But since the issue occurs rather randomly, I don't think anything but changes in the driver or madVR would be successful.

[cyberscott]

How about this one?

http://madshi.net/madVR8810test2.rar

No change. I tried to generate a new status log but it didn't generate. I'll try to get it again and post it if you need it.
Tried the log but only found this in the file... 00000000 000004d8 DLL_PROCESS_DETACH.
==
Ok, got the log file to work. Here it is. http://s000.tinyupload.com/?file_id=34300209885266799026

[shaolin95]

Guys, sorry to break the banding topic but I am trying to test for Chroma Upscaling and somehow I cant see the differences between even SoftCubic at softness 100 vs NEDDI 64 or Lanczos 4. Am I not using the best movie/images samples to see this or is something weird going on? I can see changes if I use the Image Enhancements options for example.

I am running Windows 8.1 64bits, GTX 750 with latest drivers. Latest version of madvr, ,mpc-h and lav filters. I tested this with my BenQ W1070 but also went to my other computer with an oldie gtx260 and did not see the difference either.

Maybe I am missing some settings? I want to be able to test this then move on to dithering, banding, etc but this is not a promising start lol

[huhn]

do you use DXVA native decoding? you may get DXVA chroma this way.

have you checked if your device is 4:4:4? you can test this with this: http://madshi.net/madVR/ChromaRes.png at 100% zoom.

chroma scaling has very little effect so it can be possible you simply don't see the difference because it is nearly 0.

[shaolin95]

do you use DXVA native decoding? you may get DXVA chroma this way.

have you checked if your device is 4:4:4? you can test this with this: http://madshi.net/madVR/ChromaRes.png at 100% zoom.

chroma scaling has very little effect so it can be possible you simply don't see the difference because it is nearly 0.

First of all, thank you for always helping me out man! :)

So, for the LAV video I actually have the Hardware decoder set to None so no DXVA or Cuvid selected.

I tested that image and when at 100% I can see the 4:4:4 on the right side of the screen next to the 4:2:2
I am trying to find some red fonts or something like that and get super close up to my screen to see if I can detect any changes now . :D

[vivan]

Guys, sorry to break the banding topic but I am trying to test for Chroma Upscaling and somehow I cant see the differences between even SoftCubic at softness 100 vs NEDDI 64 or Lanczos 4. Am I not using the best movie/images samples to see this or is something weird going on? I can see changes if I use the Image Enhancements options for example.Those differences are really hard to spot.
Try this sample https://dl.dropboxusercontent.com/u/16254258/android/viavr/rendering/upsampling2/upsampling2.y4m

[shaolin95]

Those differences are really hard to spot.
Try this sample https://dl.dropboxusercontent.com/u/16254258/android/viavr/rendering/upsampling2/upsampling2.y4m

Thanks a lot. Its very easy with that example indeed so I got a good idea now of how they behave. With Blurays though telling the difference is another story. :D

[andybkma]

Win7, Zoom Player, 32-bit

Hi madshi, found a bug with mVR 88.10 that is not occurring with 88.7. Whenever I try to play the attached link two AVC clips Zoom Player immediately crashes when I use either LAV or ffdshow as the AVC video decoder, whether software decoding, CUDA or Quicksync is selected. I duplicated the problem using PotPlayer and setting the same external AVC decoders as the video decoder. Oddly when I select PotPlayer's AVC built-in decoder the clips don't crash so perhaps that may help with figuring out what is wrong. Like I mentioned, this is new prob, the clips play fine with mVR 88.7. Please let me know what other info I can provide. Thanks


Download link for two clips: http://depositfiles.com/files/syv7ib458

[dansrfe]

So, overlay mode is no longer used in conjunction with D3D11? I get a lot of presentation glitches in D3D11 windowed mode and none in overlay mode.

[Asmodian]

To be fare it is impossible to get "presentation glitches" reported when in overlay mode. madVR cannot monitor overlay for something like presentation glitches and you will notice the presentation glitch line in the OSD disappears when using overlay.

Yes, overlay is a D3D9 mode and the D3D11 option overrides it.

[QBhd]

I have some more info for those of us having issues with D3D11 and PotPlayer.

I use a 72 Hz refresh rate for 23.976 playback (with ReClock) and a 60 Hz refresh rate for 29.970/59.940.... PotPlayer's onscreen info (TAB) will show 23.976-->72 (actually starts a little above 72, then falls to 72 and stays there). When the video is paused (or something else pulls focus away) there is a flicker (like going out of FSE) however ctrl-J still says FSE... and now the OSD of PP will show 23.976-->70 (and the 70 keeps falling until it hits 24)... ctrl-J shows empty queues (with no dropped frames, oddly enough) and playback gets super jittery. I have been racking my brains trying to find a way for this to work, but no setting in PP, ReClock or madVR allows me to pause the video and resume without this very very odd behavior. Playback is awesome as long as nothing interrupts it, but many times a video can be paused (can't miss anything for a beer run or to take a leak!!) and having to stop and restart each time is not an option. So I keep going back to D3D9.

To me, it seems that once the stream is interrupted, the refresh that is being reported gets altered in some way to that of the video and not the target... I just tried without ReClock, and everything I described above happens except the playback seems smooth, even with the last two queues not filling up. It's all very odd

QB

[madshi]

No change. I tried to generate a new status log but it didn't generate. I'll try to get it again and post it if you need it.
Tried the log but only found this in the file... 00000000 000004d8 DLL_PROCESS_DETACH.
==
Ok, got the log file to work. Here it is. http://s000.tinyupload.com/?file_id=34300209885266799026
Hmmmm... If you activate the debug build, you use the original v0.88.10 build, because I only sent you a modified release build. But the problem still occurs with the modified release build, too, correct? I'll have to create a debug build for that then, to get a proper new debug log.

[aufkrawall]

Could it be that downscaling in linear light can totally shift an image in an unwanted way?
Some time ago, I did some tests with downscaling game sreenshots, and with linear light everything seemed to be getting too bright.

With this image, you can notice this as well:
http://upload.wikimedia.org/wikipedia/commons/d/dc/Hermosa_beach_chip_trail.jpg

Btw: madshi, could you please teach AMD and Nvidia how to downscale properly? They are really struggling with this task with their DSR/VSR features. :( :angry:

[nevcairiel]

Could it be that downscaling in linear light can totally shift an image in an unwanted way?
Some time ago, I did some tests with downscaling game sreenshots, and with linear light everything seemed to be getting too bright.

You are using Catmull-Rom for downscaling, right? With other algorithms, linear light might not look correct, its only really recommended for CR.

[aufkrawall]

Yep, C-R. Just give it a quick try with the image above.

[James Freeman]

aufkrawall,
I think you got it wrong, the image looks darker without linear light not the other way around.

[huhn]

aufkrawall,
I think you got it wrong, the image looks darker without linear light not the other way around.

isn't this the same?

and yes LL is way brighter on this screen. i guess normal scaling is more accurate in this case.

[James Freeman]

isn't this the same?
No.

Use this image:
http://www.mediafire.com/convkey/e4a9/w7bnniib9ur3m97zg.jpg

You will clearly see that the downscaled in Linear Light keeps the visible brightness as the original image.

Look at the small detail squares (AVS HD709 written) ,white text and fine detail.
Now turn LL on and off in window mode, and go back and forth fullscreen and window with and without LL.

[huhn]

No.

Use this image:
http://www.mediafire.com/convkey/e4a9/w7bnniib9ur3m97zg.jpg

You will clearly see that the downscaled in Linear Light keeps the visible brightness as the original image.

Look at the small detail squares and the white text.
Now turn LL on and off in window mode, and go back and forth fullscreen and window with and without LL.

what i mean is.

LL scaling is bright and gamma scaling is dark is one and the same thing!

it's like this cup is half full or this cup is half empty. just a point of vision

[aufkrawall]

LL scaling is bright and gamma scaling is dark is one and the same thing!

I think that's it. Thanks.

However, the image by James Freeman is totally synthetical, whilst the image from Wikipedia is actually naturalistic.

Are there any real world images that show a real "advantage" with ll scaling?
It can definitely look too bright.

You can try this with games: Capture one screenshot in native resolution and the other e.g. with 4x higher resolution by DSR.

[cvrkuth]

have you checked if your device is 4:4:4? you can test this with this: http://madshi.net/madVR/ChromaRes.png at 100% zoom.
I did some tests too but I am a little confused. Is it normal results?

Plasma TV Panasonic TX-P42G20E RGB 4:4:4 Pixel Format PC Standard (Full RGB) in AMD CCC
According to specification RGB 4:4:4 Full RGB should be supported
http://i59.tinypic.com/a3l1tx.jpg
http://i60.tinypic.com/2s7voef.jpg

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

Samsung SyncMaster S27B350
My Samsung monitor (can't set Pixel format in AMD CCC)
http://i61.tinypic.com/2uemnly.jpg

Is it normal?

[huhn]

your panasonic TV is with it's current settings not 4:4:4.

the result is a little bit strange usually the picture gets more dark not more bright but could be easily the result of an uncalibrated screen.

supporting RGB as input doesn't mean it can be displayed same for bit deep. 12 bit input and RGB input is totally normal for all TVs but nearly all of them are not even 10 bit and a lot can't display 4:4:4 too.

[Shiandow]

aufkrawall,
I think you got it wrong, the image looks darker without linear light not the other way around.

FWIW you can actually prove mathematically that using linear light (with a gamma > 1) will make the image brighter. For more information see Jensen's inequality (https://en.wikipedia.org/wiki/Jensen%27s_inequality).

[cvrkuth]

your panasonic TV is with it's current settings not 4:4:4.

the result is a little bit strange usually the picture gets more dark not more bright but could be easily the result of an uncalibrated screen.

supporting RGB as input doesn't mean it can be displayed same for bit deep. 12 bit input and RGB input is totally normal for all TVs but nearly all of them are not even 10 bit and a lot can't display 4:4:4 too.

Thanks. Can I do something about that?

[huhn]

Thanks. Can I do something about that?

maybe there is an PC mode. i don't own an panasonic and i never calibrated one so i don't know how they usually handle this.

[cvrkuth]

maybe there is an PC mode. i don't own an panasonic and i never calibrated one so i don't know how they usually handle this.

PC and TV are connected via HDMI.
Thanks for your help!

[Qaq]

Plasma TV Panasonic TX-P42G20E RGB 4:4:4 Pixel Format PC Standard (Full RGB) in AMD CCC
According to specification RGB 4:4:4 Full RGB should be supported

Panasonic only started to support 4:4:4 without downsampling to 4:2:2 in 50, 60 series and only for incoming YCbCr. Incoming RGB always processed in YCbCr 4:2:2.
But don't try to set Catalyst to YCbCr 4:2:2, that way you will lose even more.

[cvrkuth]

Panasonic only started to support 4:4:4 without downsampling to 4:2:2 in 50, 60 series and only for incoming YCbCr. Incoming RGB always processed in YCbCr 4:2:2.
But don't try to set Catalyst to YCbCr 4:2:2, that way you will lose even more.

Maybe is subjective but on my eyes, get best picture if I choose RGB 4:4:4 Pixel Format PC Standard (Full RGB) in AMD CCC, so that Pixel Format stay.

One more question .. Should I choose 8bit instead 10bit in AMD CCC for Panasonic TV? I am not sure I can see any differences in picture quality

[Qaq]

Maybe is subjective but on my eyes, get best picture if I choose RGB 4:4:4 Pixel Format PC Standard (Full RGB) in AMD CCC, so that Pixel Format stay.
That should do best.
Should I choose 8bit instead 10bit in AMD CCC for Panasonic TV? I am not sure I can see any differences in picture quality
With my OLED the difference is day and night. With 10 bit banding is almost gone, more clarity in picture. But 10 bit takes a bit longer to switch to exclusive mode, thus if I thought I see no difference, I woudn't use it.
There were an opinions here that plazmas handle 10 bit their own way without any real benefits. Make sure you have set yours the right way. Personally I always set my TVs to PC mode for decent RGB 4:4:4 handling. That means a playback on 60Hz dispaly, but its ok for me, smooth enough. Check out your TV modes to find out the most suitable for madVR's technology and your visual perception.

[iSunrise]

One more question .. Should I choose 8bit instead 10bit in AMD CCC for Panasonic TV? I am not sure I can see any differences in picture quality
Personally, I would always do a quick test with a good grey ramp test pattern, and if you really don't see differences, then 8bit should be fine, but it's very hard to tell differences with actual movie or video content if you don't have the right samples, so a test pattern will solve that for you pretty quickly.

[cvrkuth]

That should do best.

With my OLED the difference is day and night. With 10 bit banding is almost gone, more clarity in picture. But 10 bit takes a bit longer to switch to exclusive mode, thus if I thought I see no difference, I woudn't use it.
There were an opinions here that plazmas handle 10 bit their own way without any real benefits. Make sure you have set yours the right way. Personally I always set my TVs to PC mode for decent RGB 4:4:4 handling. That means a playback on 60Hz dispaly, but its ok for me, smooth enough. Check out your TV modes to find out the most suitable for madVR's technology and your visual perception.

Personally, I would always do a quick test with a good grey ramp test pattern, and if you really don't see differences, then 8bit should be fine, but it's very hard to tell differences with actual movie or video content if you don't have the right samples, so a test pattern will solve that for you pretty quickly.
Thanks for Your help!
Regards

[dansrfe]

@madshi

Is it possible to have an option to use D39 overlay in windowed mode and D311 in exclusive mode?

I like that overlay keeps the gamma ramp for the desktop and disables it in the player window for use with the 3DLUT. Yet, it seems that I have a 10-bit display and I'd like to take advantage of that which requires D3D11.

[cyberscott]

Hmmmm... If you activate the debug build, you use the original v0.88.10 build, because I only sent you a modified release build. But the problem still occurs with the modified release build, too, correct? I'll have to create a debug build for that then, to get a proper new debug log.

Yep, same issue as before with the modified release build. Would it be useful to send you a debug log of 88.8 to compare?

[madshi]

Yep, same issue as before with the modified release build. Would it be useful to send you a debug log of 88.8 to compare?
No. But please try these two (both debug builds):

http://madshi.net/madVR8810test3.zip

Please let me know if either of them works (or both). If either of them does *not* work, please upload a debug log for that. If both don't work, please upload a debug log for both. Thanks.

[cyberscott]

No. But please try these two (both debug builds):

http://madshi.net/madVR8810test3.zip

Please let me know if either of them works (or both). If either of them does *not* work, please upload a debug log for that. If both don't work, please upload a debug log for both. Thanks.
Neither of those worked. One slight difference I noticed was with the present queue is now stuck at 2-5/8 instead of the previous 1-4/8 on both builds. Here are the logs.
http://s000.tinyupload.com/?file_id=37081719334988377767

[XRyche]

I know this is the madVR thread but since Shiandow's debanding algorithm is being tested along side madVR's, I hope the thread will forgive me for the following.
In my tinkering with the deband algorithm's this weekend I have confirmed for myself a rather happy alternative use for Shiandow's algorithm set at 1.00 strength and .01 margin with grain. It does an excellent job of deblocking and denoising my old dvr rips at sub 480p resolution without totally killing detail (in tandem with nnedi3) and making it a fuzzy mess. Well more fuzzy than it already was.
This is just my opinion, but I think it would really be beneficial (for me and possibly even a small minority of users) if Shiandow might be able to optimize his debanding algorithm further down that path and possibly madshi could use it as a denoiser/general deblocker for madVR somewhere in the near future.
Not every user uses madVR for just relatively clean HD content. I for one have a lot of old dvr files that I just can't see myself getting rid of and no amount of recoding that I can personally do will improve the quality of these files.
This is just a suggestion an by no means a demand in any way, shape or form.

[MysteryX]

madVR v0.88.10

* fixed: some problems when using NNEDI3
* slightly modified "high" debanding preset, once more

I tried NNEDI3 again and it still doesn't work. What's interesting is that if I apply NNEDI3 frame quadrupling, it does change rendering times, but that's because it applies Jinc3 twice. See attached picture. I'm using dual-graphics Radeon HD 7670M on Windows 7 x64.

http://s15.postimg.org/shofpg4qv/NNEDI3.jpg (http://postimg.org/image/shofpg4qv/)

[huhn]

have you checked it without profiles?

[Asmodian]

The double "> Jinc3 AR" report is normal if you weren't using doubling but were using upscaling refinement with the "refine the image after every ~2x upscaling step" option checked. It looks like NNEDI3 doubling is unavailable or that profile isn't the one being used?

The bold text doesn't change once you open the settings page so changes to the source (probably not the issue here) or the settings can cause madVR to use one of the other profiles without any indication.

[tobindac]

Call me crazy but my favorite feature after the possibility for 10bit, was that the uninstaller/installer no longer does "DING!" everytime is run.

[Blackfyre]

Call me crazy but my favorite feature after the possibility for 10bit, was that the uninstaller/installer no longer does "DING!" everytime is run.

I actually miss the "DING" :p:p:p

[MysteryX]

The double "> Jinc3 AR" report is normal if you weren't using doubling but were using upscaling refinement with the "refine the image after every ~2x upscaling step" option checked. It looks like NNEDI3 doubling is unavailable or that profile isn't the one being used?

The bold text doesn't change once you open the settings page so changes to the source (probably not the issue here) or the settings can cause madVR to use one of the other profiles without any indication.
I was using upscaling refinement, but with the other option checked to do it only once. If I disable refinement, the double upscaling is gone.

It's using the right profile. NNEDI3 never worked on my computer, no matter the driver version, no matter if I reinstall Windows.

However... Jinc4 and Lanczos8 were removed, probably because NNEDI3 is a better alternative. But since I don't have NNEDI3, there are cases where Jinc4 and Lanczos8 serve me, so for now I'm better to stick to the old version.

[MSL_DK]

With v0.88.9 and v0.88.10 Average stats > rendering never stabilize, it jumps up and down and i've have frame drops. It applies to Jinc and jumps from 6ms to 22ms. If I choose Bicubic it settles at around 18ms after a few seconds. If you need more information about my sys and settings, say so.

[madshi]

Neither of those worked. One slight difference I noticed was with the present queue is now stuck at 2-5/8 instead of the previous 1-4/8 on both builds. Here are the logs.
http://s000.tinyupload.com/?file_id=37081719334988377767
The first log makes sense, but the 2nd doesn't. The 2nd log indicates that it wasn't from the build 2 I intended you to test. Either I uploaded the wrong file, or you tested with the wrong file. To be safe, I've uploaded the 2nd build again:

http://madshi.net/madVR8810test4.rar

Could you please test this build again? Thanks!