madVR - high quality video renderer (GPU assisted)

[Anime Viewer]

I'm getting lockups after updating to the latest version of madvr. it was either madvr or strongene hevc decoder i recently installed. disabled strongene and lockups still occur. playing hevc with lav, ffdshow raw, & svp on intel gpu.

The usual questions asked when people report lockup/freezing:

Did you try resetting madVR to default settings, and see if the problem still exists?

Have you scanned your system for corrupt/mislinked filters/codec using something like the codec tweak tool? If not do so, fix any errors it finds, and to be on the safe side you can also tell it to associate default codec/filters. Sometimes disabling decoders/filters/codec doesn't really stop them, so you may have to uninstall the strongene decoder you reported installing.

[THX-UltraII]

video chain question, again......

Is this the best (and even more imporant: correct) video chain when my main purpose is Blu-Ray movie playback:

Blu-Ray .m2ts file
Device properties in madVR on PC levels (0-255)
the native display bit depth on 10 bit (or higher)
LAV filters video settings: RGB output on untouched (as is)
display device (Epson TW9200 projector) on EXTENDED

[THX-UltraII]

Now that (I think) that I know how all things works I ll try to summarize the BTB/WTW discussion. You basically have 3 options:

1.
Set madvr to pc levels, set gpu to pc levels, set TV to pc levels (0-255). This means, that BTB and WTW will be clipped, it also means that there is expansion involved, which might lead to banding. madvr dithering does a good job of eliminating that.

2.
Set madvr to TV levels, set GPU to pc levels, set TV to TV levels. This means there will be neither clipping nor expanding.

3.
If your TV doesn't have settings to change black levels and is fixed to TV levels: Set madvr to PC, GPU to TV. BTB and WTW will be clipped, and there is double processions going on (expansion by madvr -> reduction by gpu).

The second option is the best quality wise, because there is no processing of the levels involved, and therefore no banding introduced. The problem is that if you use your PC for anything but viewing through madvr, your levels will be messed up.

There is also a 4th option if you want both limited(video) and full-range(pc applications) to display properly. If you can adjust input mapping on your display (usually low or limited vs. normal or full) then use option 2 and switch display input to normal when displaying desktop applications.I found my own post from a few years ago. Does this still apply and is it correct how I stated it back then?

[huhn]

I found my own post from a few years ago. Does this still apply and is it correct how I stated it back then?

you should clipping everything below 16 and everything over 235 that's simply how is should work. not clipping it will destroy the CR.

best is pure PC level if your end device supports PC level properly.


i only heard that some panasonic supports PC level but the results are bad with banding for what ever reason.

[Barnahadnagy]

Interestingly, my ST50 doesn't have banding issues at PC levels.
About chroma scaling: There are very few scenes / movies where I can actually see a difference. One of these is the following example (http://screenshotcomparison.com/comparison/133624), where the chroma coding is atrocious (don't ask how it turned out to be like that). I think this demonstrates the potential in Bilateral quite well. Interestingly SuperChromaRes makes about no difference here (And neither does Bicubic vs Jinc, softcubic does make aliasing better but at 100 it starts to bleed out of edges too).

[ryrynz]

Interestingly, my ST50 doesn't have banding issues at PC levels.

Neither does my VT50, PC,TV or TV,PC.. look pretty much exactly the same on the greyscale ramp.

[toniash]

Anyone knows how to solve this:
"error: input is not a PCH file: 'cl_kernel.h.pch'
fatal error: 'cl_kernel.h.pch' does not appear to be a precompiled header file" ??
I get it when trying to use s-xbr on XP SP3 and Nvidia 352.86
A driver reinstallation solved it

[THEAST]

I reported an issue a while back with duplicate frames being displayed after "excessive jumping" in MPC-HC with madVR used as renderer but nobody else seemed to have experienced it. Now I have pretty much pinpointed the issue: it is caused by madVR's smooth motion and for me it is 100% reproducible with AVI and WMV files (doesn't seem to happen with MP4 and certainly doesn't happen with MKV). When I jump through a video multiple times consecutively using MPC-HC's built-in jump backward/forward shortcut keys and then stop, sometimes after less than a second some, frames are displayed that are clearly from a different scene. Disabling smooth motion eliminates the issue regardless of container. This doesn't seem to be decoder/splitter-dependent since I can reproduce it with every combination of ffdshow/LAV as decoder and Halli/LAV as splitter. This also isn't hardware-dependent since I experience this issue both on my laptop with GTX 850M and my desktop with HD7950. OS is Win 7 SP1 x64 with madVR and MPC-HC x86.

Steps to reproduce:
1. Play a WMV file with madVR used as renderer and smooth motion set to "always".
2. Start jumping back and forth in the file using MPC-HC's built-in shortcuts 3-5 times and wait for one second or two between each jump (jumping forward seems to work best).
3. Stop jumping and watch the video carefully.
4. After 1-2 seconds you should see some random irrelevant frames on the screen that last less than one second and might repeat again after a few seconds.

[James Freeman]

I found my own post from a few years ago. Does this still apply and is it correct how I stated it back then?

Yes, it does.
Personally, I use #2.

[digby]

The usual questions asked when people report lockup/freezing:

Did you try resetting madVR to default settings, and see if the problem still exists?

Have you scanned your system for corrupt/mislinked filters/codec using something like the codec tweak tool? If not do so, fix any errors it finds, and to be on the safe side you can also tell it to associate default codec/filters. Sometimes disabling decoders/filters/codec doesn't really stop them, so you may have to uninstall the strongene decoder you reported installing.

did the reset and use the tweak tool to set some splitters to lav - doubt it matters since i use mpc. still locked up tho'. the picture freezes but the sound continues for another 5-10 secs before silence. mouse still moves but everything is frozen. on another computer with amd gpu, there are no lockups; just this one with the intel gpu. i wonder if it has anything to do with svp - i'm using madvr on the intel gpu and svp on a 5450. i've been using this split setup for months tho' so why would it be a problem now?

[SecurityBunny]

Is debanding suppose to be so performance intensive? With it off, my rendering time is 10ms. With it on high/high (or even low/high), rendering time is at 30-31ms. Enabling 'don't analyze gradient angles for debanding' drops the rendering time to 20ms.

MadVR 0.88.14 x64
MPC-HC 1.7.9.30 x64
GeForce 353.38

[Warner306]

Is debanding suppose to be so performance intensive? With it off, my rendering time is 10ms. With it on high/high (or even low/high), rendering time is at 30-31ms. Enabling 'don't analyze gradient angles for debanding' drops the rendering time to 20ms.

MadVR 0.88.14 x64
MPC-HC 1.7.9.30 x64
GeForce 353.38

On a mid-level to low-level graphics card I would consider debanding to be a medium processing feature compared to other madVR features.

[SecurityBunny]

On a mid-level to low-level graphics card I would consider debanding to be a medium processing feature compared to other madVR features.

I don't think my GTX 780 classified is considered a mid/low level graphics card. Lately my rendering times have been all over the place, much higher than normal. I think madVR might be putting my card in a lower clock state since the last few updates. Is there any way to stop it from downclocking the card after a minute of playback besides running it at full power constantly?

Also, what is the recommended hardware decoder to use with madVR in LAV video settings?

[edigee]

DXVA2(native). Or NVIDIA CUVID. DXVA2(copy-back) I only use for 10bit h265 60Hz videos ,which is not the case for your card because it doesn't have full hardware decoding for h265/HEVC.
I have GTX 960.
NVIDIA CUVID is working the same as DXVA2(native) in terms of CPU usage(very low on both) but somehow it gives weird colors ,for instance skin color tones are a bit reddish and less detailed.
Edit:For Kepler cards 347.88 is still the best driver. The last drivers seems to work good(well ,some of them) only for Maxwell cards. I have a gt 640 on another rig and the latest drivers(since 350.12) are causing issues of all kind.

[aufkrawall]

I don't think my GTX 780 classified is considered a mid/low level graphics card. Lately my rendering times have been all over the place, much higher than normal. I think madVR might be putting my card in a lower clock state since the last few updates. Is there any way to stop it from downclocking the card after a minute of playback besides running it at full power constantly?

Yes, set power profile to maximum performance in the Nvidia control panel (but just for MPC HC, not globally). Then you won't have maximum clock either, but GPU should stay in boost state.
But why would you want to do this? The GPU should choose a higher clock fast enough.


Also, what is the recommended hardware decoder to use with madVR in LAV video settings?
There aren't any drawbacks over CPU when using DXVA2 CB, LAV is doing this very efficiently.
However, maybe you don't want to enable it for 4k since the Kepler VPU isn't fast enough for 4k with high bitrates.

[huhn]

DXVA2(native). Or NVIDIA CUVID. DXVA2(copy-back) I only use for 10bit h265 60Hz videos ,which is not the case for your card because it doesn't have full hardware decoding for h265/HEVC.
I have GTX 960.
NVIDIA CUVID is working the same as DXVA2(native) in terms of CPU usage(very low on both) but somehow it gives weird colors ,for instance skin color tones are a bit reddish and less detailed.
Edit:For Kepler cards 347.88 is still the best driver. The last drivers seems to work good(well ,some of them) only for Maxwell cards. I have a gt 640 on another rig and the latest drivers(since 350.12) are causing issues of all kind.


cuvid does technically the same as dxva copyback it just force the GPU in high powerstate which totally beats the reason to use a harware decoder and to be honest i think CUVID is totally worthless these days. and the reddish would be a bug do you have a screen with OSD ?

[SecurityBunny]

Yes, set power profile to maximum performance in the Nvidia control panel (but just for MPC HC, not globally). Then you won't have maximum clock either, but GPU should stay in boost state.
But why would you want to do this? The GPU should choose a higher clock fast enough.

Unfortunately that is the exact thing I would like to avoid. Ideally I'd like to keep power usage and temperature down, not run the card at maximum power for the duration of an entire video. I've been aiming for a rendering time under the vsync interval for smooth playback. When I first start a video, rendering time is 8ms with the GPU usage clocked at ~20% at 1,110mhz. After a minute or so playing the content, rendering time becomes 18ms with GPU usage spiked to ~45% at 666mhz.

I don't recall having this problem with the GPU throttling a few months back. :scared:

DXVA2(native). Or NVIDIA CUVID.
NVIDIA CUVID is working the same as DXVA2(native) in terms of CPU usage(very low on both) but somehow it gives weird colors ,for instance skin color tones are a bit reddish and less detailed.



There aren't any drawbacks over CPU when using DXVA2 CB, LAV is doing this very efficiently.
However, maybe you don't want to enable it for 4k since the Kepler VPU isn't fast enough for 4k with high bitrates.

cuvid does technically the same as dxva copyback it just force the GPU in high powerstate which totally beats the reason to use a harware decoder and to be honest i think CUVID is totally worthless these days. and the reddish would be a bug do you have a screen with OSD ?

Thanks. I've been using no hardware decoder for years since I read somewhere that it wasn't recommended to be used with madvr. If there isn't a problem with it decoding 10bit encoding nowadays, I'll go ahead and enable DXVA2 (native) if that is the best option out of the three for quality and speed.

Is there much of a difference between native and copy-back?


Edit:For Kepler cards 347.88 is still the best driver. The last drivers seems to work good(well ,some of them) only for Maxwell cards. I have a gt 640 on another rig and the latest drivers(since 350.12) are causing issues of all kind.

Unfortunately I can not downgrade driver versions since I am on Windows 10. It automatically forces an update so I'm stuck on 353.38 for the time being. Fortunately I haven't ran into any problems with it.


DXVA2(copy-back) I only use for 10bit h265 60Hz videos ,which is not the case for your card because it doesn't have full hardware decoding for h265/HEVC.
I have GTX 960.

I'm able to play h265/HEVC content. I'm assuming this is possible because it is falling back to the software decoder and that full hardware decoding only provides the benefit of being faster?

One final question. Is there a list of what 'trade quality for performance' options are safe to check that have little to no impact on the actual quality?

[michkrol]

Unfortunately that is the exact thing I would like to avoid. Ideally I'd like to keep power usage and temperature down, not run the card at maximum power for the duration of an entire video. I've been aiming for a rendering time under the vsync interval for smooth playback.
You need to aim below frame interval, not vsync.
If you don't get any dropped frames, I don't see how throttling would be an issue. I get almost 32ms render times with powersaving and the playback is rock solid with 24fps videos, smoothmotion enabled.
I'm able to play h265/HEVC content. I'm assuming this is possible because it is falling back to the software decoder and that full hardware decoding only provides the benefit of being faster?
If you're using LAVFilters (also MPC-HC's built-in codecs), the decoder falls back to software for formats that your hardware can't decode, that includes 10bit videos on almost all GPUs.

[huhn]

Thanks. I've been using no hardware decoder for years since I read somewhere that it wasn't recommended to be used with madvr. If there isn't a problem with it decoding 10bit encoding nowadays, I'll go ahead and enable DXVA2 (native) if that is the best option out of the three for quality and speed.

all h264/h265 decoder should have the same quality or the decoder is buggy.
software is still the best way and most save way to decode videos it simply has better error handling.

hybrid h265 decoding looks like a bad joke. i wouldn't use it.

Is there much of a difference between native and copy-back?

native has some limitation and copyback avoids these limitations.
the performance impact from copyback on nvidia system is totally meaningless copyback is just more flexible.

[madshi]

Yes, it's the AR filter which leads to the darkening.
But without, the ringing is very annoying. The ringing filter of Jinc doesn't have this issue.

Btw: Lanczos upscaling looks terrible with the cartoon sample. Ringing outta hell...
The next build will have super-xbr chroma upscaling either with the strict/agressive AR or with the high quality (but slower) madVR AR algo. So please try again with the next build (out in a couple of hours).

Perhaps that may explain why present queue doesn't fill all the way initially when using 24hz, but it doesn't explain the bug. The bug occurs on all refresh rates as far as my testing goes.

The bug is when toggling to exclusive fullscreen, then windowed, then back to fullscreen again, the rendering and present queue completely drop and the playback stutters. This only occurs with D3D11 10bit. You need to pause the video and unpause it to get the queues to fill properly again. Using D3D9 10bit or D3D11 8bit, I can freely toggle in and out of exclusive fullscreen without queues not filling.

Plus the fact it takes longer going fullscreen with D3D11 10bit than compared to D3D9 10bit/D3D11 8bit.
Maybe I can see something if you create a debug log from this situation. Try to reproduce the problem quickly, so the log doesn't get too long. Then when the problem occurs, let it stutter for 10 seconds, do not change anything at that point (this is important! otherwise the log will be hard to interpret), then simply close the media player.

TYVM for providing xbr-25, I didn't have time yet but I'll recompare it all again tomorrow. This said I was plenty happy with NEDI+SR in 88.13 and I'm afraid that the EE in xbr and NNEDI3 is just part of their design, I mean you want sharper edges you got them duh.......NEDI looks more natural and seems to be less agressively seeking edges, different methods for differents needs IMO. NEDI is perfect for 720@1080p, xbr50 for tiny videos but ideally I would like to find an in-between for untouched DVD's that require something sharp but not overly.
So I can remove super-xbr-25 again?

Please post screenshots that show where/why you like NEDI better than super-xbr. Thanks.

I assume that NNEDI3 will never be really replaced in terms of quality (if quality is defined as the absence of aliasing and ringing and cleanest reconstruction of lines).
Don't give up hope just yet. I mean I've no idea, maybe you're right. Or maybe not, we'll see...

Hello, sorry if thats naive question but how about implementing into madvr option of dynamically set a scaling algorithm basing on gpu usage.
Maybe at some time in the future, but not any time soon. Too many other things to do first.

-I still far prefer NEDI+SR for 720p@1080p because it looks natural and just really great IMHO :)
Screenshots? :)

-Actually I'm cool with the current stock settings of SR, I mean 2 passes don't seem to be enough and 4 too much, 0.75 seems to hit the spot not being too soft or too sharp, 0.10 softness looks equally good because 0.05 is edgier and 0.15 too blurry and HQ disabled looks a hell lot better too yay! When I enable HQ the "watching through a window" feeling disappears, it looks dull and sorta noisy....really hate the look, major bottleneck at work here......"tout ça pour ça" :scared:
Sorry to say, but the default settings are sharpness 1.0 and softness 0.0. Probably you still had the old defaults from the old SuperRes algo stored. Currently plan is to remove softness. Also, currently HQ enabled seems to be better in several cases, although maybe worse in some others. But current tendency goes in favor of HQ enabled.

If you have examples of where HQ looks worse, please show screenshots - thanks!

I fully agree that if anything's currently lacking in mVR it's an sxbr strength knob or at least more steps because 37 might be right up my alley and 87 yours :o
Not sure who you're agreeing with here, maybe yourself? But sorry, no knob planned for now.

I have some confusion about the fact that there is now sharpening and Super res in 3 different places. If I am using one for upscaling and one for native res do they cross each other? For instance if I tick sharpening under Image Enhancements does it also affect chroma upscaling and/or upscaling refinement? Which order should I be using them in for which settings?
Consider the current builds as a work in progress in terms of sharpening and SuperRes. We're still trying to figure out the optimal parameters and stuff. So nobody right now can tell you exactly which is the "optimal" setting.

Also. Since there is now a sharpening option can we have a luma denoiser (with sizer and strength) in future builds?
Sure, if you write one?

My specs
i5 2500k with gtx 960 , display aoc e2352phz
Why i'm getting dx3d 8bit?
& what color output i should use in nvidia setting? YCbCr or RGB in Ful range???
What should i use for my display 8bpc or 12 bpc?
i watch anime frequently & is my setting ok for it
Can anyone suggest best setting for gtx 960 OC ,i5 2500k,8 GB RAM.
I'm new to this things
These questions are being asked a lot. Most users here are tired answering the same question again every week, so that's why you didn't get many (or any?) replies yet. Might make sense to look for some madVR guides. There are several of them, some better and some worse.

FWIW, you could simply try using the default settings and check if they work alright. If they do, you can play with the settings (e.g. image upscaling or doubling) and check if your GPU can handle them and if you can see a difference or not.

Why is it I get 0 issues during playback on my 60Hz 1080p HDTV, but I do get crazy number of presentation glitches on my 1080p 120Hz monitor, using identical settings? Average stats rendering time is 15ms and present time is below 6-8 or so in both cases. Max stats rendering time often switches between 16ms and 33ms and present time usually stays at about 33ms, but at times goes to 140ms. I get these presentation glitches only in Exclusive mode, although I tried to disable "present several frames in advance", but that made no difference. Turning off Direct3D 11 made no difference. Enabling/Disabling "present a frame for every Vsync" also made no difference. I also tried to tone down settings to make it easier on GPU (no NNEDI3, no SuperRes, etc), but that also made no difference!
Do you have smooth motion FRC enabled? If so try to disable it.

OS? GPU?

Agree. NEDI+SR combine very well together. Jinc+SR doesn't work well. I like that XBR has a very low performance cost compared to other algorithms, but it doesn't do nearly as good as NEDI+SR.
Screenshots?

Madshi you should write a madVR benchmark tool like SVP has. Then all the users can upload their results to a public spreadsheet. It would help troubleshoot performance problems by allowing comparisions and give a general rule of what to expect with a certain gpu/machine with different settings.
madVR is changing fast so it isn't time for standard benchmark tools yet. :)
^

Just wanted to say that the latest version is absolutely astonishing!
The image quality has improved visibly and I am using the same settings as always, no issues whatsoever.

Madshi is there a way I can donate something for this project? Do you need or accept donation?
Thanks. At the moment I don't accept donations just yet. Plan to start doing that when madVR reaches v1.0, which is still some time away...

Yes, super-xbr has improved a lot in the latest version. It may replace Jinc now, but still in some cases I find Jinc a tad better due to a more natural look. Please open this image (http://www.mediafire.com/view/wewuu3et8pdg8xt/sample.JPG) and toggle between Jinc and super-xbr. Avatar movie can be a good sample too.
That's a 4K image. Doubling that with super-xbr results in 8K. On my 1680x1050 LCD display I've no idea where to look in that image for differences. Can you create a screenshot comparison, with maybe a part extracted where the difference is especially strong in favor of Jinc?

I'm getting lockups after updating to the latest version of madvr. it was either madvr or strongene hevc decoder i recently installed. disabled strongene and lockups still occur. playing hevc with lav, ffdshow raw, & svp on intel gpu.
Is the media player itself also locked up (e.g. menu doesn't open, buttons don't even move when being pressed etc)?

Is debanding suppose to be so performance intensive? With it off, my rendering time is 10ms. With it on high/high (or even low/high), rendering time is at 30-31ms. Enabling 'don't analyze gradient angles for debanding' drops the rendering time to 20ms.
Debanding with gradient angle analyzation is relatively computation heavy. The old madVR builds didn't analyze gradient angles for the "high" debanding preset. The latest build does. Maybe you're used to the old "high" preset, which of course was faster? You can get it back simply by checking the "don't analyze gradient angles" tweak. Of course the debanding vs detail loss ratio improves if you do let madVR analyze the gradient angles.

I don't think my GTX 780 classified is considered a mid/low level graphics card. Lately my rendering times have been all over the place, much higher than normal. I think madVR might be putting my card in a lower clock state since the last few updates.
Possible. I had implemented a few smaller performance improvements. And "high" deband got slower, see above. The performance improvements were really small, though, nothing dramatic (other than super-xbr and FineSharp).

[6233638]

Hi there,
the latest Calman release now has some improvements when running Calman on the same PC as madTPG. The changes have been inspired by your "complaint"... :D So it would be awesome if you could try the new build and maybe provide feedback here:

http://www.spectracal.com/forum/viewtopic.php?f=94&t=5729

Stacey Spears has actively asked for feedback about this. Would be great if you could tell him/us if it works better now, or if there's still something that could be improved.Thank you. This generally seems to be working, though it doesn't seem to be entering FSE mode, so the LUT is only going to be based off 8-bit measurements rather than 10-bit.

[Anime Viewer]

If there isn't a problem with it decoding 10bit encoding nowadays, I'll go ahead and enable DXVA2 (native) if that is the best option out of the three for quality and speed.


Some extremely old (poorly coded) videos can show pixelation with hardware acceleration on, but usually those are rare to come across these days. To a degree some of the madVR error correction features may lessen or eliminate some of the pixelation, but its hard to say for sure. Try with it enabled, and if you see pixelation in your videos then you can set it to none instead. Native should be the fastest given that it doesn't copy data back out to the system/cpu. Depending on the speed of your non-gpu components (memory/cpu) you may see/notice a difference between copyback and native, or you may not. Bottom line is in most cases native should be faster. This is more of a LAV than madVR topic. There are quite a bit of articles/posts about native vs copyback you can find with a search engine, and if those don't address the issue to your liking you can post in the LAV forum.

[aufkrawall]

Some extremely old (poorly coded) videos can show pixelation with hardware acceleration on
With copyback? How is this possible?

Btw: There were copyback options in madVR some time ago. What happened to them?

[madshi]

Thank you. This generally seems to be working, though it doesn't seem to be entering FSE mode, so the LUT is only going to be based off 8-bit measurements rather than 10-bit.
Would you mind reporting this directly in the spectracal thread (link see my post which you quoted)? I mean I can play messenger and duplicate your posts there, and Stacey's posts here, but it would make the communication between the two of you very slow, and cost me additional time. Thanks!

Btw: There were copyback options in madVR some time ago. What happened to them?
They got deleted because the latest LAV build has faster copyback routines than madVR ever had. If you want to use DXVA decoding with the best quality, use the LAV copyback functionality. If you want DXVA decoding with the fastest speed, use DXVA native, but then you'll get ever so slightly lower chroma quality with some GPUs (Intel, NVidia).

[aufkrawall]

Alright, thanks.
Is OpenCL used on AMD to prevent quality loss?

[madshi]

No. OpenCL interop cost on AMD is too high for OpenCL to be of use for things like that.

[omarank]

That's a 4K image. Doubling that with super-xbr results in 8K. On my 1680x1050 LCD display I've no idea where to look in that image for differences. Can you create a screenshot comparison, with maybe a part extracted where the difference is especially strong in favor of Jinc?
I was talking about chroma upscaling, not image doubling. I also mentioned it in my post. Can you please check again or do you still need the screenshot comparison to be created?


Feedback on SuperRes:

Questions:

1) Which image upscaling/doubling algorithm do you like to use with SuperRes and why?
2) Which values do you like for "strength" and "softness". Please note that the default values are strength=1.0, and softness=0.0. And these *may* be the best values. But you can still try other values to see if you like them more.
3) The "use HQ downscaling" option changes the overall "look" of SuperRes a bit. Which look do you prefer? Please note that with the option turned on, you may have to increase the number of passes, because the SuperRes effect is slightly less intense with this option turned on.
4) How many passes should be used as default?

1) NNEDI3 because of the least aliasing and ringing. It looks the cleanest.
2) The default values look good to me. I tried other values but could not decide whether it’s getting better or worse.
3) I prefer HQ downscaling to be enabled.
4) 2 passes look fine. However, I set the number of passes to 3. There is not much difference though.

[madshi]

I was talking about chroma upscaling, not image doubling. I also mentioned it in my post. Can you please check again or do you still need the screenshot comparison to be created?
It's still a 4K image. On a quick check I didn't know where to look. I did a quick check but didn't see any obvious differences between Jinc or super-xbr chroma upscaling.

Feedback on SuperRes:

1) NNEDI3 because of the least aliasing and ringing. It looks the cleanest.
2) The default values look good to me. I tried other values but could not decide whether it’s getting better or worse.
3) I prefer HQ downscaling to be enabled.
4) 2 passes look fine. However, I set the number of passes to 3. There is not much difference though.
Thanks! You were using strength=1.0 and softness=0.0, right? I'm asking just to be sure, because madVR did store and reuse the settings you used from the older SuperRes.

[edigee]

cuvid does technically the same as dxva copyback it just force the GPU in high powerstate which totally beats the reason to use a harware decoder and to be honest i think CUVID is totally worthless these days. and the reddish would be a bug do you have a screen with OSD ?
You're right. Last time I saw those kind of difference was on a GT 640 and older versions of madVR. It seems there is OK now.

[THX-UltraII]

It s been a while since I asked this question but are there ANY plans at all in the future to get madVR work with PowerDVD? I would LOVE to use a 3Dlut file with powerDVD!!!

[madshi]

madVR v0.88.15 released

http://madshi.net/madVR.zip

* super-xbr image doubling AR optimization: 20% performance boost
* super-xbr chroma upscaling now supports higher quality AR algo -> 6% slower
* super-xbr chroma "AR" option now switches between low/high AR quality
* modified Bilateral chroma upscaling algorithm parameters
Now image doubling with super-xbr is faster than doubling with Jinc AR!

For super-xbr chroma upscaling, v0.88.14 either supported "no AR" or the original super-xbr strict AR. Now v0.88.15 doesn't support "no AR", anymore. Instead the "activate anti-ringing filter" option now activates the slower high-quality madVR AR algorithm. With the "activate anti-ringing filter" option disabled, the strict AR algorithm is used instead. I know, this is slightly confusing, but I think it's more useful to offer these two different AR algos instead of offering no AR at all, considering that the strict AR is so fast that it barely makes a performance difference. I've thought about which AR algo to use when the AR option in madVR is activated. But my thinking was that users usually expect higher quality and slower performance when that option is enabled, so that's how I decided that with the option activated, the slower high-quality madVR AR algo is used, and with the option deactivated, the strict AR algo is used instead.

FWIW, at this point I think the stream of super-xbr tweaks from the previous weeks is at an end. No dramatic super-xbr changes planned for the near future, anymore.

Would still love to get more feedback about the new SuperRes algo.

[kasper93]

Thanks for new release.

For some reason madshi.net domain is on Malvertising filter list by Disconnect (https://s3.amazonaws.com/lists.disconnect.me/simple_malvertising.txt) (apparently uBlock addon uses this list). You might want to contact them to clear this up.

[Ver Greeneyes]

Thanks for the new release! What AR algorithm is used for super-xbr image doubling? Or does AR not make sense for that?

[madshi]

Thanks for new release.

For some reason madshi.net domain is on Malvertising filter list by Disconnect (https://s3.amazonaws.com/lists.disconnect.me/simple_malvertising.txt) (apparently uBlock addon uses this list). You might want to contact them to clear this up.
Would you mind complaing to uBlock and/or Disconnect? I've already reported this false positive to them a full month ago, and they still haven't done anything about it. Honestly, Disconnect seems to suck big time. I've never had any company take even remotely this long to fix a false positive. The ironic thing is that my site has never had any advertizing, ever! So how could it possibly be added to some "malvertising" list??

Thanks for the new release! What AR algorithm is used for super-xbr image doubling? Or does AR not make sense for that?
Image doubling uses the high quality madVR AR algorithm.

[Ver Greeneyes]

Image doubling uses the high quality madVR AR algorithm.Thanks! That makes sense.

[aufkrawall]

The new AR filter for super-xbr is better than the original one, colors get better seperated with it instead of worse.
However, I still find the ringing more obvious than with Jinc AR, unless I set sharpness to 50. But this causes less clear color contrast again.
So I personally will stick to Jinc for chroma with its aspect of "positive ringing".

Have the values for SuperRes for chroma changed? I can't remember it worsening ringing with default values, but "now" it does.

NNEDI3 luma, eternal love.. :p :D

Regarding SupeRes for image doubling: I'm still sticking to passes 1, strength 0.3, softness 0 and HQ downscaling.
Seems like a no brainer for me to use it at least this way since I never could see it making anything look worse.

Bilateral chroma upscaling is still terrible.

[SecurityBunny]

Maybe I can see something if you create a debug log from this situation. Try to reproduce the problem quickly, so the log doesn't get too long. Then when the problem occurs, let it stutter for 10 seconds, do not change anything at that point (this is important! otherwise the log will be hard to interpret), then simply close the media player.


Here you go.

Went from windowed (1s) -> fullscreen (1s) -> windowed (1s) -> fullscreen (~10s) -> alt + f4 to close player in fullscreen. Queues not filling occurred, as well as playback appearing to stutter but not showing dropped frames.

https://www.dropbox.com/s/hi8yiom7xcwayc3/madVR%20-%20log.txt?dl=0

Not sure if it would make much of a difference, but here is another debug where I waited for queues to completely fill before switching modes. The bug occurred in both attempts and is easily repeatable by just toggling between windowed and fullscreen exclusive mode twice.

Windowed (3s) -> fullscreen (3s) -> windowed (3s) -> fullscreen (~10s) -> alt + f4.

https://www.dropbox.com/s/cqo5yif5vykomad/madVR%20-%20log2.txt?dl=0

Again as a reminder, this only seems to occur with D3D11 10bit. D3D9 10bit and D3D11 8bit both seem to be fine.

In a completely unrelated situation, it seems unchecking fullscreen exclusive mode to use the windowed fullscreen path crashes the nvidia display driver immediately when attempting to go fullscreen. Not sure if it is a problem with Windows 10 build 10164, GeForce 353.38, or madVR - but have been experiencing this for a few months. I primarily use fullscreen exclusive for the 10bit output so it isn't a big concern for me, just thought I'd let you know about it.

[huhn]

It s been a while since I asked this question but are there ANY plans at all in the future to get madVR work with PowerDVD? I would LOVE to use a 3Dlut file with powerDVD!!!


what would you get from powerdvd you can't get from anydvd hd if you want to use madVR?

[THX-UltraII]

what would you get from powerdvd you can't get from anydvd hd if you want to use madVR?Hi Huhn, PowerDVD (and TMT) are the only HTPC software player out there that can play frame packed 3D Blu-Ray movies.

[huhn]

but even with madVR in these programs. madVR still can't do frame packed 3d.

[Anima123]

So I can remove super-xbr-25 again?

Madshi, please keep super-xbr-25 option, which could be useful for some not well coded videos. super-xbr-25 would be sharp enough but not overly sharped.

For Chroma upscaling, shouldn't we use a less sharper algorithm? In this case, super-xbr-25 can also be useful.

Anyone ever tried with super-xbr-25 for chroma upscaling?

[Catclaw]

Any idea why using CUVID or NNEDI3 chroma upscaling crashes windows 10? and is any progress being made towards madvr on windows 10? :D

[ryrynz]

Any idea why using CUVID or NNEDI3 chroma upscaling crashes windows 10? and is any progress being made towards madvr on windows 10? :D
Until it's released its almost pointless looking at it. The code changes daily..

[viewer]

i prefer super-xbr strict AR. super-xbr high-quality AR is something ringing.:scared:

http://s5d1.turboimagehost.com/t/23504075_super-xbr150AR_madVR0.88.14.png (http://www.turboimagehost.com/p/23504075/super-xbr150AR_madVR0.88.14.png.html)
http://s5d1.turboimagehost.com/t/23504076_super-xbr150HQAR_madVR0.88.15.png (http://www.turboimagehost.com/p/23504076/super-xbr150HQAR_madVR0.88.15.png.html)

[xabregas]

Hi,

I must say super-xbr is amazing, but why not put it in luma upscaling below jinc? I found every luma upscaling algorithm very weak compared to luma doubling even for 720p > 1080p upscaling. Even knowing it will upscale 2 x 720p and then downscale to 1080p using bicubic or wtv i prefer using luma double because i feel it has better PQ to my eyes. So, why not put that option in luma upscaling? It shouldnt be much gpu intensive given the performance boost we get from NNEDI3 to super-xbr. Meaning, using super-xbr to 720p -> 1080p upscale.

[ryrynz]

i prefer super-xbr strict AR. super-xbr high-quality AR is something ringing.:scared:

Woah, that's messed up! Look at all those green lines.. Madshi, any room for improvement there?

[michkrol]

* modified Bilateral chroma upscaling algorithm parameters
Thanks for the new version of madVR, especially this.

After some testing, Bilateral is now my chroma upscaling algorithm for anime.
It's pure magic, especially with high quality sources :D

i prefer super-xbr strict AR. super-xbr high-quality AR is something ringing.:scared:
You should give Bilateral a try. I really like the results with high quality sources.
Your sample with bilateral chroma scaling below.

http://s5d3.turboimagehost.com/t/23506577_4024565.png (http://www.turboimagehost.com/p/23506577/4024565.png.html)
Hi,

I must say super-xbr is amazing, but why not put it in luma upscaling below jinc? I found every luma upscaling algorithm very weak compared to luma doubling even for 720p > 1080p upscaling. Even knowing it will upscale 2 x 720p and then downscale to 1080p using bicubic or wtv i prefer using luma double because i feel it has better PQ to my eyes. So, why not put that option in luma upscaling? It shouldnt be much gpu intensive given the performance boost we get from NNEDI3 to super-xbr. Meaning, using super-xbr to 720p -> 1080p upscale.
I don't understand you reasoning behind this.

Biggest questions: what would madVR do if you selected s-xbr upscaling and nnedi3 doubling? What image scaling algo would madVR use after doubling with s-xbr?
It's under doubling because you can only double, not scale by any (arbitrary) ratio, performance has nothing to do with this. The up-/down-scaling algo used after doubling is important for image quality.

[aufkrawall]

Any idea why using CUVID or NNEDI3 chroma upscaling crashes windows 10? and is any progress being made towards madvr on windows 10? :D


Cuvid + needi3 chroma scaling is broken here (GTX 980, 353.38, Windows 10 10162 x64) too, but NNEDI3 chroma scaling only works well (e.g. with DXVA2 CB).
madVR is running very well here on Windows 10 apart from that unimportant Cuvid issue.

[Vyral]

After some testing, Bilateral is now my chroma upscaling algorithm for anime.
It's pure magic, especially with high quality sources :D


Really !?

I didn't even try this algorithm because many users here said it was horrible.

[aufkrawall]

With lowres sources, heavy color bleeding can be observed.

Jinc AR:
http://abload.de/thumb/jincwso6v.png (http://abload.de/image.php?img=jincwso6v.png)

Bileteral:
http://abload.de/thumb/bilnkp4w.png (http://abload.de/image.php?img=bilnkp4w.png)