madVR - high quality video renderer (GPU assisted)

[seiyafan]

I really appreciate the performance data you have captured in an easily to digest form. Your dedication is to be commended regardless of public recognition (or not). Others have attempted to simplify the MadVR option choices based on pure performance (objectivity) but eventually became frustrated with the ongoing complexity (never ending exceptions) of the task, not to mention the subjectivity involved in making those choices.

My next upgrade will be a '4K' display - most likely a Sony 55" XBR-55X850A. I currently am using a Sony KDL 40W3000 (for the past 6 years through out numerous PC upgrades) and have been *very* satisfied by how well it has served as an arms-length 'PC monitor'. I've never viewed 'TV' on it... I firmly believe 4K will be of more use to me than most folks intending to use it for 'entertainment' purposes.

Therefore I view your AMD/MadVR/MPC-HC findings with great interest. Don't get discouraged if you think your hard work is going unrecognized - you are learning a hellava lot about how it all plays together (or doesn't) in your bench-marking exercise. Thanks again :-)

Glad I could help!

Just a word of caution to everyone that no one should view those numbers as absolute, the main purpose was to compare the performance difference between different cards. With the same setting I got different rendering times with H.264, H.265 and bluray videos, sometimes the difference is quite big (like 35%). I think it has to do with encoding but I am not 100% sure.

[MistahBonzai]

...Just a word of caution to everyone that no one should view those numbers as absolute, the main purpose was to compare the performance difference between different cards...

Understood - in the end it's all 'relative'. :)

[cyberbeing]

"Delay playback start" unfortunately is not bug free. I even have a couple bugs [1] (http://bugs.madshi.net/view.php?id=104) [2] (http://bugs.madshi.net/view.php?id=5) opened about other quirks with it.Are those bugs still there when tested with the latest versions?

Hmm... It looks like Issue #5 (http://bugs.madshi.net/view.php?id=5) with DirectShowSource seeking may have been resolved by coincidence in 0.87.5, as it was still reproducible in 0.87.4.

Issue #104 (http://bugs.madshi.net/view.php?id=104) I'm still able to reproduce in 0.87.5.

That 'bug'(?) remains repeatable with madVR v0.87.5.
Since I don't watch any PAL content, I just workaround the problem by setting a special rule to disable Smooth Motion with 25fps, which LAV Filters always reports for still images.

if (srcFPS = 25)

[ryrynz]

I've actually encountered a number of times on my HTPC where the playback just pauses like you describe in 104. I'm not toggling the subtitles either.
It doesn't happen very often, I thought it might have been a media player/network streaming issue.. but perhaps it's not.

[6ari8]

Thanks for the update!


I'm not sure if this is only happening to me, but when I minimize the player that was on my secondary monitor (extended), regardless of whether it was paused or playing, the player freezes and it has a black screen when brought back and I'm unable to close it without ending the process from the task manager. I've tested with both the latest MPC-BE and MPC-HC.


If the player was on the main monitor, no freeze occurs.


EDIT: Forgot to say that XySubfilter isn't loaded, different combinations of rendering>general settings and different dithering settings give the same result.

EDIT2:

I shouldve mentioned that I use ReClock.

When I use ReClock, the issue mentioned previously occurs.

When not using ReClock, the freeze doesn't occur but when I bring back the minimized window, the first frame that appears is strange:

http://imageshack.com/a/img138/5719/mmhr.png


It stays for less than a second then it goes back to normal. When I kept minimizing and maximizing, It mostly happens but sometimes it doesn't and sometimes it's replaced by a black frame. The number characters keep changing too. After many minimizing and maximizing, that frame doesn't appear and is replaced by a black frame.


Using:
MPC-BE/MPC-HC
LAV Fitlers 0.61.0
XySubfilter 3.1.0.682 (happense whether loaded or not)
Reclock 1.8.8.1
GeForce GT 650M, latest drivers (334.89)
Windows 7 SP1.

[James Freeman]

* fixed: #136: when playback is stopped, madVR now always turns the video off

Finally...
Also, when closing the video completely (Ctrl+C) in MPC-HC it clears the memory completely unlike before.
I'll say it again, this is a great build.

Thank you.


Thanks - I had been using 1.3.10 (MPC LumaSharpen Shader), 1.4.1 works just fine - I have it enabled as a "post resize shader".
BTW: Anyone expecting obvious razor sharp results need to look elsewhere - this shader is very subtle using the defaults when applied to good quality 720P being upsized to 1080P.

Maybe with default settings (they are very subtle).
But it can be extremely (ridiculously/unwatchable) sharp too.


#define sharp_strength 0.65
#define sharp_clamp 0.035


#define sharp_clamp = set the threshold of where the sharpening begins to take place.
The higher it is (max 1.0) the more the finer detail will get sharpened too (may lead to halo/ringing artifacts around edges).
With lower setting, only the "blurry" textures will get sharpened, leaving the finer (already sharp) detail untouched, thus preventing halo/ringing artifacts.

#define sharp_strength = Try 100 (with clamp 1.0).
The image will be so sharp it'll cut you... ;)

[burfadel]

With version 0.87.5, when I am playing a video on a secondary screen, when it goes to the next (or previous) video, it crashes. I tried all different setting etc. to try and narrow down the cause, and also reset to default values, but nothing worked.

I then went back to 0.87.4 and it worked flawlessly as before (even when changing back to all the options I had before. I then copied over the files from MadVRFinalDither4, and it still worked flawlessly. I then tried MadVRyuvFullRange and that worked flawlessly too. Tried 0.87.5 again, crash! It is only 0.87.5 that doesn't work, so something changed between the final test build and the final version to cause this issue. It ONLY occurs when playing on the secondary screen.

[turbojet]

madshi: thanks for the new build, the blank screen overlay toggle issue seems fixed, also resizing overlay window is much smoother.

Try this if you want sharper lumasharpen:
#define sharp_strength 1.0
#define pattern 3
#define sharp_clamp 0.5

Pattern 3 does a much better job of limiting artifacts at higher strengths then 8 or any of the others. The 17 taps seems to help, I wish I knew how to use even more taps. You can raise strength if you really want sharp but I settled on 1.0 since it looked over sharpened above it.

Using it pre resize really helps sharpen fine details, but you should use lanczos resize with it the other resizers work against it. Using it pre and post resize isn't that big of a hit on gpu.

This really needs it's own thread the author, ceejay.dk, was going to make one.

[Shii]

After updating to madVR 0.87.5 and xySubFilter to latest 3.1.0.682 Beta2 Zoom Player no longer loads subtitles via xysubfilter with Smart Play. Is any way to force loading subtitle filter without rolling back to 0.87.4?

[James Freeman]

Pattern 3 does a much better job of limiting artifacts at higher strengths then 8 or any of the others.
Thanks, for the heads up.


Using it pre resize really helps sharpen fine details.

In lower res Pre resize works better because the size of the pixel is 1:1, effectively its just a small image in the middle in the native resolution of your screen, so the edge is pixel sharp for the algorithm to process (like it should be).

but you should use lanczos resize with it the other resizers work against it.

Lanczos naturally sharpens the image and creates ringing as it goes, like most sharp upscalers.
The LumaSharpen shader does its thing without creating additional ringing because of the Threshold parameter.
Maybe when using this shader Pre resize its better to use a softer upscaler in madVR that does not create ringing, thus getting the best of both worlds.

This deserves another thread, so no more posting about it here.
I'll do some testing and report in a new thread, I think I'll call it "Image Sharpening & MadVR".

[madshi]

Oh, so the dither pattern didn't actually change every frame when smooth motion was enabled?
It changes for the smooth motion frc blended frames, too, but smooth motion frc does not always render at the display refresh rate. It only renders at the fps necessary to achieve the desired effect. E.g. for 24fps@60Hz, it renders less than 60fps.

Here's the result of R9 290X

This is the result of upscaling to 1440 with a H.265 encoded video, rendering time is 10-30% longer with bluray, ouch! If you are upscaling to 4k expect the time to be even longer. I also tried to "maxed out" MadVR with 256 neurons in both luma and chroma doubling, as you can see, it answered my question of why madshi decided to stop at 256 and not go higher. :P
Good comparison! The algorithm itself stops at 256, that's why I'm not offering more neurons. FWIW, if you want to use luma doubling I'd suggest to reduce the other settings. Chroma upscaling doesn't have to use NNEDI3. And you don't need Jinc3 for image upscaling, if you already double luma. The effects on the final image quality of NNEDI3 chroma upscaling or Jinc3 image upscaling are rather small, compared to doubling luma with NNEDI3. So I'd suggest to test with BicubicAR chroma upscaling, BicubicAR (or maybe Lanczos3AR) image upscaling. Also I don't recommend using chroma doubling at all. I've just added it to the madVR settings dialog for people who like to play with "mad" settings.

So basically what I'd find more interesting would be to test how fast the various GPUs out there are with BicubicAR for chroma + image upscaling + image downscaling, chroma doubling disabled and luma doubling enabled. This for HD content -> 1440p, for HD content -> 4K, and for SD -> 1440p and SD -> 4K. For SD content luma quadrupling might also be interesting to test. For luma quadrupling 16 neurons should be good enough, or max 32 neurons. For luma doubling more neurons should help image quality (slightly), for luma quadrupling more neurons are probably a bit wasted.

Does "use OpenCL to process DXVA NV12 surfaces(Intel, AMD)" do anything? In my test it reduced rendering time from 38.36ms to 38.24ms.
Yes, it should change things slightly, with Intel and AMD GPUs. Whether it improves or slows down performance, I don't know.

Impending Doom (http://blogs.msdn.com/b/directx/p/directx12.aspx) :devil:
I guess Microsoft felt it necessary to answer AMD's Mantle...

I seem to be getting a madvr 0.87.5 crash (it then sends a report - when I can fill out the details before zoom player crashes) and then next time its zoom player that crashes. When I am full screen and click the control bar 'Next track' when at the end of a file. Or use my mouse shortcut button.

Just seems to be whilst in full screen with this new 0.87.5. I'd never seen the madvr crash reporter before today. Very rarely have zoom player crashes, never mind madvr crashes.

Reverting back to default release 0.87.4 fixes the issue.

0.87.5, either clicking next track or letting the video change to the next video on its own whilst in full screen this seems to either crash madvr and brings up the crash reporter, but it then stops the crash reporter as I'm going through its steps, with a zoom player crash. So I rarely get a chance to fill out the madvr crash reporter.

I did lower the cpu queue to 20 and gpu queue to 12 - no effect on the crashing. But reverting to 0.87.4 does fix things as I mentioned.
I've received 2 crash reports from you, and I replied to one of them. The cause of the crash in your case seems to be that you're running out of RAM. I don't know why.

Anybody seeing RAM consumption problems with v0.87.5, compared to older versions?

I'm not sure if this is only happening to me, but when I minimize the player that was on my secondary monitor (extended), regardless of whether it was paused or playing, the player freezes and it has a black screen when brought back and I'm unable to close it without ending the process from the task manager. I've tested with both the latest MPC-BE and MPC-HC.
Is this a new problem with v0.87.5, or did this occur with older versions, too? Does this only happen on the secondary monitor, or also on the primary?

With version 0.87.5, when I am playing a video on a secondary screen, when it goes to the next (or previous) video, it crashes.
Do you get the madVR crash box? If so, please send the crash report to me.

After updating to madVR 0.87.5 and xySubFilter to latest 3.1.0.682 Beta2 Zoom Player no longer loads subtitles via xysubfilter with Smart Play. Is any way to force loading subtitle filter without rolling back to 0.87.4?
You need to deregister and reregister XySubFilter. If that doesn't help, this is an issue you should report to the XySubFilter thread.

Using it pre resize really helps sharpen fine details, but you should use lanczos resize with it the other resizers work against it.
If the source itself is soft, maybe running it before resize can make sense to sharpen the source. However, if you just want to remove the softness caused by upscaling process itself, it makes more sense to sharpen after upscaling.

[6ari8]

Is this a new problem with v0.87.5, or did this occur with older versions, too?

Yes, this is a new problem. It didn't occur with older versions before.

Does this only happen on the secondary monitor, or also on the primary?

It only happens on the secondary monitor.

[James Freeman]

DirectX 12
I guess Microsoft felt it necessary to answer AMD's Mantle...
If it is, Great.

[burfadel]

Do you get the madVR crash box? If so, please send the crash report to me.

Unfortunately no, it's mpc-hc that crashes, but is caused by 0.87.5 (but with none of the test builds like I said).
Yes, this is a new problem. It didn't occur with older versions before.

It only happens on the secondary monitor.

Same here. The problem is only when I play stuff on the secondary monitor, when I play it on my main screen it works perfectly. Maybe the cause is not entirely unrelated to the issue I am having.

[markanini]

Sharing my Intel HD3000 settings for interlaced SD:
DXVA luma scaling in MadVR, Yadif 25/30Hz mode and no hardware decoding in LAV Video decoder. The output totally smooth and reasonably sharp. Any other settings makes the HD 3000 drop frames.

[Vyral]

Thanks for this new version madshi.
I had a strange behavior yesterday, just after I updated from v0.87.4 to v0.87.5.

- v0.87.4 - Jinc 3 AR / Jinc 3 AR / Catmull-Rom AR-LL ; Debanding ; smooth motion ; random dithering
1 frame repeated every 42 sec

- v0.87.5 - Jinc 3 AR / Jinc 3 AR / Catmull-Rom AR-LL ; Debanding ; smooth motion ; ordered dithering
1 frame repeated every 5 sec

- v0.87.5 - Jinc 3 AR / Jinc 3 AR / Catmull-Rom AR-LL ; smooth motion ; ordered dithering
1 frame repeated every 42 sec

No impact on GPU load (~31%) or rendering time (13ms).
Tested with a 720p 24fps video.

EDIT : More testing and the problem is gone. Ignore this post.

[Q-the-STORM]

Here (https://www.mediafire.com/?7tbms5fxmqaes4e) I've uploaded a modified nv_disp.inf (English) & nv_dispi.inf (International) x64 which adds support for GTX 780 Ti to 327.23.

Thanks, it is now allowing to install... but now i get in device manager "Windows has stopped this device because it has reported problems. (Code 43)"

any way to fix it?

[annovif]

Madshi, last year i bought a 7750 for jinc3-4, now i have to buy a 280x for NNEDI3 and others features . Can i know your roadmap for 2014, so if i have to buy a 290x i buy it directly:)

Thank you, Fabio

[nevcairiel]

NVIDIA drivers use a unified architecture, so any driver which at least supports Kepler (and maybe even prior) should work on a GTX 780 Ti with an inf mod.

While they use a unified driver for all architectures, it does not mean that old drivers generally support newer cards.
This seems rather risky to do if you value stability. :p

[Werewolfy]

I have an issue with the latest build v0.87.5. When I start a video in Overlay mode I have a black screen, only the sound is working. The picture appears when I resize the video or hit Ctrl+J for example.

It happens only in Overlay Mode and it works fine on madVR v0.87.4 or finaldither4.

Windows 8 Pro 64 bits
Geforce GTX660 327.23

Edit : I had the idea to test FSE mode while writing this message. I activated FSE mode and then desactivated it and now Overlay mode works fine. I really don't know why, I don't change anything else.

I post this message anyway, it might help someone.

[cyberbeing]

Thanks, it is now allowing to install... but now i get in device manager "Windows has stopped this device because it has reported problems. (Code 43)"

any way to fix it?

Maybe that driver really is a no-go if it's failing to load after a clean install. I'm surprised, especially considering the GTX780 Ti uses 100% identical install parameters to the GTX780 & Titan in later drivers. I guess you could try one of the Quadro maintenance releases which adds support for the latest models like K6000 (Quadro variant of GTX780 Ti) and bugfixes. If you are running Windows 8.1, you may be out of luck though. And on Win8.0, I've heard you need to install non-WHQL driver in Test Signing mode. I personally run Windows 7.

Quadro 321.10 (December 17, 2013) (http://www.nvidia.co.uk/download/driverResults.aspx/71564/en-uk) + Modifed Inf for GTX 780 Ti (https://www.mediafire.com/?vrbwgb4wo4m9rao)
I was using this driver for awhile on my GTX770, and can confirm it works with madVR OpenCL.

While they use a unified driver for all architectures, it does not mean that old drivers generally support newer cards.
This seems rather risky to do if you value stability. :p
It depends. As long as there is something similar, it will usually work. My GTX770 for example is functional with all Kepler drivers, even those long before its release date. I've also been using Quadro specific driver branches off and on for years without issue.

[Q-the-STORM]

I guess you could try one of the Quadro maintenance releases which adds support for the latest models like K6000 (Quadro variant of GTX780 Ti) and bugfixes. If you are running Windows 8.1, you may be out of luck though. And on Win8.0, I've heard you need to install non-WHQL driver in Test Signing mode. I personally run Windows 7.

Quadro 321.10 (December 17, 2013) (http://www.nvidia.co.uk/download/driverResults.aspx/71564/en-uk) + Modifed Inf for GTX 780 Ti (https://www.mediafire.com/?vrbwgb4wo4m9rao)
I was using this driver for awhile on my GTX770, and can confirm it works with madVR OpenCL.


I run win7 pro x64....
and this driver works with 780 Ti, thanks... now I can finally use NNEDI3 :P

[cyberbeing]

I run win7 pro x64....
and this driver works with 780 Ti, thanks... now I can finally use NNEDI3 :P

Happy to hear it worked this time :)

I have to admit, I'm fond of this recent trend by NVIDIA to do maintenance releases of their older driver branches for a prolonged period of time.

[StinDaWg]

FWIW, I can do luma doubling for 24p content just fine with my HD7770, which should be much slower than your GPU. But this is only with 24p content. For 60p luma doubling my GPU is too slow.
How are you able to do this? I also have a 7770, and even with all the relevant trade for performance checked, and bicubic (no AR) for chroma, CR for downsizing, I can't run any of the nnedi luma doubling with 720p24 content on a 1080p screen. 16 neurons gives me ~42 ms rendering time and 64 neurons gives me ~62.

This is my card...
http://www.newegg.com/Product/Product.aspx?Item=N82E16814202084

[nevcairiel]

I have to admit, I'm fond of this recent trend by NVIDIA to do maintenance releases of their older driver branches for a prolonged period of time.

Thats not necessarily a recent trend, but its usually limited to Quadro releases which always require manual hacking to get them to install, so its quite a bit annoying...

[cyberbeing]

I'm pretty sure it is a recent trend, or at least them classifying them as out-of-cycle maintenance releases for a specific branch number. That appears to have started with R304 last year.

Back when I owned my 7800GTX 512, I used to track NVIDIA driver branches before it was widely known. You can probably find old posts from me on Guru3D forums from 2006 or so, trying to explain to people how NVIDIA's driver branch system worked. ;)

[secvensor]

HD7850DC22GD5V2
1080p 60fps 10bit
none drop frames
settings:
http://i33.fastpic.ru/thumb/2014/0306/28/98b32f10ab895f86b7fe976660da8928.jpeg (http://fastpic.ru/view/33/2014/0306/98b32f10ab895f86b7fe976660da8928.jpg.html)

But these two modes don't work:
http://i60.fastpic.ru/big/2014/0306/c0/de158cc475b2b8bde8289fca31e436c0.jpg

[mzso]

Is there a way to tell whether Error diffusion is working other than increased CPU usage? Technically my GTS450 is DX11 capable so I'm gonna assume it works.


If this problem occurs with v0.87.5, then please let me know. Otherwise please ignore this problem for now.
The problem is near non-existent with v0.87.5. I get 20MB plus for 10 playback cycles. So it's around 40 times less.
http://abload.de/thumb/pot-madvr-leak4mmrit.png (http://abload.de/img/pot-madvr-leak4mmrit.png)

[leeperry]

settings:
http://i33.fastpic.ru/thumb/2014/0306/28/98b32f10ab895f86b7fe976660da8928.jpeg (http://fastpic.ru/view/33/2014/0306/98b32f10ab895f86b7fe976660da8928.jpg.html)
Your settings are pretty much all wrong IMHO, I'd reset mVR completely at this point.

[secvensor]

leeperry
Then give me the correct settings.

[seiyafan]

HD7850DC22GD5V2
1080p 60fps 10bit
none drop frames
settings:
http://i33.fastpic.ru/thumb/2014/0306/28/98b32f10ab895f86b7fe976660da8928.jpeg (http://fastpic.ru/view/33/2014/0306/98b32f10ab895f86b7fe976660da8928.jpg.html)


What's the output resolution? I suspect it's playing 1080p movie on a 1080p monitor/TV, so upscaling is not used at all. Otherwise please see my test result of 290X with that setting on page 1218.

[secvensor]

What's the output resolution? I suspect it's playing 1080p movie on a 1080p monitor/TV, so upscaling is not used at all. Otherwise please see my test result of 290X with that setting on page 1218.

1920x1448@75mhz
1600x1200@100mhz

[Shiandow]

HD7850DC22GD5V2
1080p 60fps 10bit
none drop frames
settings:
http://i33.fastpic.ru/thumb/2014/0306/28/98b32f10ab895f86b7fe976660da8928.jpeg (http://fastpic.ru/view/33/2014/0306/98b32f10ab895f86b7fe976660da8928.jpg.html)


I fear that NNEDI3 also doesn't work, I think no GPU is able to actually run with those settings. Your "trade quality for performance" screen also still has options from v0.85.4 which shouldn't be there. At this point you may want to try reinstalling MadVR completely and updating to the latest AMD drivers. I'd also recommend resetting your settings, since some of your settings are impossible and in the other cases I prefer the default.

[secvensor]

Shiandow
I have simply changed the name of the version 0.85.4 and have added last point at the very bottom.
original:
http://i58.fastpic.ru/thumb/2014/0306/cf/504c58343e4f6085f664819a4743ffcf.jpeg (http://fastpic.ru/view/58/2014/0306/504c58343e4f6085f664819a4743ffcf.jpg.html)

[pirlouy]

@secvensor: are you using these settings for benchmarks, or do you use it as a base ?
It looks like you're trying to put every options to use the most CPU/GPU possible, instead of searching for stability. There are settings you should not change if you don't understand them well.

[huhn]

@secvensor

what are you trying to say us what's the point?

you watch a 1080p file on a 1080p screen and post image doubling settings no gpu can handle and they are not even active?

256 neuron for chroma no gpu can do this for a 23p 1080p 4:2:0 source..

[Francois76l]

Good work Madshi! As usual!
The new dithering options are working well ;)

Nothing to deal with the new version but can someone explain to me what is vsync interval?
On 4 differents monitors with a 23.976Hz movie, my vsync interval is the same: 20.85ms. Is it normal?

[secvensor]

pirlouy
I use it as a base on 1600x1200@100mhz
end as test on 1920x1448@75mhz.
0 dropped frames on any size (320p~1080p).
huhn
if i set "several frames at advance", my gpu begins drope frames.
http://i60.fastpic.ru/thumb/2014/0306/05/0739304bbfe2f45484a9be3f38054f05.jpeg (http://fastpic.ru/view/60/2014/0306/0739304bbfe2f45484a9be3f38054f05.jpg.html) http://i60.fastpic.ru/thumb/2014/0306/8a/09de1fa33d59dc3545afe9caf0cd288a.jpeg (http://fastpic.ru/view/60/2014/0306/09de1fa33d59dc3545afe9caf0cd288a.jpg.html)

[ryrynz]

My experience was when I reset my settings to default most if not all of my set options remained the same, I thought there was a bug because the default for Luma doubling wasn't set at 32.

I had to reset to defaults a second time for everything to reset properly.

[madshi]

With version 0.87.5, when I am playing a video on a secondary screen, when it goes to the next (or previous) video, it crashes.
I was able to reproduce this on the secondary monitor. Should be fixed in v0.87.6.

I'm not sure if this is only happening to me, but when I minimize the player that was on my secondary monitor (extended), regardless of whether it was paused or playing, the player freezes and it has a black screen when brought back and I'm unable to close it without ending the process from the task manager. I've tested with both the latest MPC-BE and MPC-HC.
I was not able to reproduce this problem here. But there are a couple of other bug fixes in v0.87.6, maybe one of them fixes this problem for you, too?

Madshi, last year i bought a 7750 for jinc3-4, now i have to buy a 280x for NNEDI3 and others features . Can i know your roadmap for 2014, so if i have to buy a 290x i buy it directly:)
I'm sorry, but I don't have a roadmap. Not even a secret one. I do what I do when I do it. I do not like to plan a long time into the future, if I don't have to.

I have an issue with the latest build v0.87.5. When I start a video in Overlay mode I have a black screen, only the sound is working. The picture appears when I resize the video or hit Ctrl+J for example.
Yes, I saw this, too. Should be fixed in v0.87.6.

How are you able to do this? I also have a 7770, and even with all the relevant trade for performance checked, and bicubic (no AR) for chroma, CR for downsizing, I can't run any of the nnedi luma doubling with 720p24 content on a 1080p screen. 16 neurons gives me ~42 ms rendering time and 64 neurons gives me ~62.
Well, 42ms is pretty close. 1000/23.976 = 41.71ms. If you can just shove off 2-3ms it would probably run fine, in fullscreen exclusive mode at least. I've been testing this with a 1680x1050 monitor, that probably explains why it worked for me.

Is there a way to tell whether Error diffusion is working other than increased CPU usage? Technically my GTS450 is DX11 capable so I'm gonna assume it works.
If you can't see a difference then be happy and stick to ordered dithering. It's quite a bit faster... ;) If you want to see a difference, try setting your display to 3 bit in the madVR device setup. Just for kicks, only, of course.

Nothing to deal with the new version but can someone explain to me what is vsync interval?
It's the duration of one refresh period (VSync) of your display. If you have 20.85ms that suggest that your display is set to (1000 / 20.85 =) 47.96Hz.

[madshi]

madVR v0.87.6 released

http://madshi.net/madVR.zip

* fixed: #090: FSE mode switched to 23Hz instead of 24Hz in Windows 8
* fixed: #127: crash when jumping to next video file on secondary monitor
* fixed: #173: overlay: exiting multiple windows in same order -> black screen
* madTPG now forces ordered dither, but you can en/disable colored & dynamic
* added support for new subtitle API ISubRenderConsumer2::Clear()
To be honest, I hadn't planned to release a new official build so quickly, but I've fixed two important bugs (one crash, one overlay related) and found one very important improvement. So I thought it would be worth a new build. The improvement is that FSE mode should hopefully work with the correct refresh rates now even in Windows 8 and 8.1. So you can finally use 23.976Hz, 24.000Hz, 59.940Hz and 60.000Hz again, and madVR should happily switch to exactly the refresh rate you want. At least it seems to work on my AMD GPU, while all older builds didn't. I think it should work for NVidia and Intel, too.

(Of course the refresh rate problem was - and still is - a serious bug in Windows 8. I've just hacked my way around it now.)

[6233638]

24Hz works correctly for me on Nvidia now, and also seems to have fixed another bug (http://bugs.madshi.net/view.php?id=105) too.


Looking back, I should have done a better job documenting how to reproduce that one - I think I had a detailed post about it here, but that was 8 months ago.

[JonnyRedHed]

All looks good to me so far :) thanks.

[madshi]

24Hz works correctly for me on Nvidia now, and also seems to have fixed another bug (http://bugs.madshi.net/view.php?id=105) too.
All looks good to me so far :) thanks.
Yey! :)

[iSunrise]

No issue here. Are you sure it's not this bug (http://bugs.madshi.net/view.php?id=142) I reported which is causing your issue?
Indeed, it is related to smooth motion. If I disable it, still images at least show again. However, if I let the still image on for a couple of seconds, it suddenly disappears and I have a black screen again. It seems madVR 0.87.5 for some reason empties the queues after a couple of seconds, even after both keep being filled with 0.87.4. I have just reproduced that behaviour 3 times. Not even pausing the playback helps in this situation.

So there definitely is some change in behaviour from 0.87.4, which breaks still images for me.

I used these settings:
(1) enable "delay playback start until render queue is full"
(2) disable "enable smooth motion frame rate conversion"

Otherwise still images don´t work at all.

[huhn]

(Of course the refresh rate problem was - and still is - a serious bug in Windows 8. I've just hacked my way around it now.) hacker these days now i can at least think about to use windows i hope this works!

http://bugs.madshi.net/view.php?id=174
i can't reproduce this error with fullscreen exclusive mode on 87.6
and after down draging to 87.5 i got a hard time to reproduce this but in the end is got it again, then it happens over and over again but 87.6 looks fine so who cares.

pirlouy
I use it as a base on 1600x1200@100mhz
end as test on 1920x1448@75mhz.
0 dropped frames on any size (320p~1080p).
huhn
if i set "several frames at advance", my gpu begins drope frames.
http://i60.fastpic.ru/thumb/2014/0306/05/0739304bbfe2f45484a9be3f38054f05.jpeg (http://fastpic.ru/view/60/2014/0306/0739304bbfe2f45484a9be3f38054f05.jpg.html) http://i60.fastpic.ru/thumb/2014/0306/8a/09de1fa33d59dc3545afe9caf0cd288a.jpeg (http://fastpic.ru/view/60/2014/0306/09de1fa33d59dc3545afe9caf0cd288a.jpg.html)

both screens use window mode both screen didn't show any problems at all.

[burfadel]

Thanks! 0.87.6 works fine, no secondary screen crash bug as stated in the changelog :)

[Werewolfy]

madVR v0.87.6 released

http://madshi.net/madVR.zip

* fixed: #090: FSE mode switched to 23Hz instead of 24Hz in Windows 8
* fixed: #127: crash when jumping to next video file on secondary monitor
* fixed: #173: overlay: exiting multiple windows in same order -> black screen
* madTPG now forces ordered dither, but you can en/disable colored & dynamic
* added support for new subtitle API ISubRenderConsumer2::Clear()
To be honest, I hadn't planned to release a new official build so quickly, but I've fixed two important bugs (one crash, one overlay related) and found one very important improvement. So I thought it would be worth a new build. The improvement is that FSE mode should hopefully work with the correct refresh rates now even in Windows 8 and 8.1. So you can finally use 23.976Hz, 24.000Hz, 59.940Hz and 60.000Hz again, and madVR should happily switch to exactly the refresh rate you want. At least it seems to work on my AMD GPU, while all older builds didn't. I think it should work for NVidia and Intel, too.

(Of course the refresh rate problem was - and still is - a serious bug in Windows 8. I've just hacked my way around it now.)

Thanks for the new build but your workaround for FSE mode in Windows 8 doesn't work for me. It still swiths to 23.976Hz instead of 24.000Hz. The only difference I can see is when my screen is set to 60.000Hz and when I go in fullscreen, for a few seconds madVR tells my screen is in 0.000Hz and I have a lot of stutters. After a few seconds, everything works fine but it's set at 23.976Hz.

I'm under Windows 8 Pro 64 bits and I have a Geforce GTX660 327.23

[leeperry]

Oh great, you're working on the W8 refresh rate rolling problem. Hopefully you'll iron it out for good and I'll be able to update to the next W8 iteration with mVR able to go 24/50/60Hz without a itch :)

[6ari8]

I was not able to reproduce this problem here. But there are a couple of other bug fixes in v0.87.6, maybe one of them fixes this problem for you, too?


I just tried v0.87.6 and the freezing problem is gone. Thanks!

Though, this issue still is there but it's not affecting my experience and doesn't happen with neither overlay or FSE mode:

when I bring back the minimized window, the first frame that appears is strange:

http://imageshack.com/a/img138/5719/mmhr.png

It stays for less than a second then it goes back to normal. When I kept minimizing and maximizing, It mostly happens but sometimes it doesn't and sometimes it's replaced by a black frame. The number characters keep changing too. After many minimizing and maximizing, that frame doesn't appear and is replaced by a black frame.


I think this might be a problem on my side as I'm starting to feel I need a clean Windows reinstall because I haven't done that in over a year while I installed and uninstalled a lot of stuff that might've broke something throughout. I also think that the registry may be cluttered and need to be reset.