madVR - high quality video renderer (GPU assisted)

[nevcairiel]

2) Does "crop black bars" have any side effects? Should it be always ticked?

It can crop away hard-coded subtitles inside the black bars or TV station logos or stuff like that.

[pirlouy]

madshi can you share your future plans for madVR?
He said several times he does not want to speak about his plans.
I though you followed madVR development for many years...

[James Freeman]

He said several times he does not want to speak about his plans.
I though you followed madVR development for many years...
Ah yes, thanks for reminding me.
I'm so eager for the next thing that I forget all about patience, sorry about that madshi. :D

Move along.

[Razoola]

It seems madVR is causing MPC-hc to crash if NVidia 3D vision is enabled when going into fullscreen? If I set MPC-HC to use EVR then it does not crash. Is there any way I can make a meaning full log for you to track this one given madvr does not seem to crash?

[dimitrik]

Has anyone else experienced an MPC-HC crash on every exit for the last several minor versions of madVR? Only way to close the crash reporter window is killing the process in task manager. Windows 7 64-bit, 32-bit MPC-HC/madVR, GTX 970. I can give more details if there isn't a known crash that behaves in this way. I haven't experimented with settings lately to find if there's a particular cause because I've assumed the crash would go away after the next update, but many updates later I'm starting to wonder if it isn't a known issue.


I've had a similar problem with MPCHC hanging on every exit but the crash reporter window doesn't come up.

I fixed it by cleaning out all codecs and reinstalling, upon which time I switched to the x64 version of MPCHC.

Unfortunately I had to redo all the settings but there was no way around it.

[YGPMOLE]

This is probably a GPU driver issue, maybe in combination with D3D9. You could try if D3D11 improves the situation. Or try enabling the madVR display mode switcher and entering those refresh rates you want madVR to use. Then madVR will try its best to force Direct3D to use the correct refresh rates.

The D3D11, the madvr display mode switcher and the MPC-HC fullscreen changer (and the various combination between them) didn't fix the problem. Now I'm trying a clean install after a format and I'll let you know.

The new built fixed the subtitle positioning: thanks for your continuous and generous work!!!

[Weirdo]

Hope you can find the time to add support for the MP4 rotation flag. :)

[lanzorg]

Yes you can. Here (https://github.com/sebdelsol/pyPlayer/blob/master/miniplayer.py)is a movie player written in Python that uses madVR's COM interfaces.

There's a catch though. You won't be able to access a madVR object that has been created in another process, directly controlling madVR in MPC-HC from an external Python script for instance.
I've tried... to no avail, and ended up writing a whole player :)
I guess it might be possible with some weird dll injection.

Oh that's awesome but a little bit complicated for what I want to do.
Could you share python code sample for enabling smooth motion?
That would help me a lot.

[JarrettH]

Hi madshi or anyone, I have a video without black bars where the movie resolution is 960x720 and the target rectangle is 1280x720 (viewing in window)

In a window, it shows chroma > super-xbr and luma > jinc...does this mean the video is already being upscaled? I don't understand why it doesn't just show a 960x720 window.

In full screen it shows chroma > super-xbr, image x > super-xbr and image y > super-xbr

madvr 0.88.21

[nevcairiel]

Hey madshi or anyone, I have a video without black bars where the movie resolution is 960x720 and the target rectangle is 1280x720 (viewing in window)

In a window, it shows chroma > super-xbr (normal) and luma > jinc...does this mean the video is already being upscaled? I don't understand why it doesn't just show a 960x720 window

Sounds like it needs aspect ratio correction to 16:9 (ie. 960x720 -> 1280x720), which uses scaling, of course.

[YGPMOLE]

Now I'm trying a clean install after a format and I'll let you know.

The switching issue is gone!!! After a clean installation of Windows 10, madVR it's rock solid again even in Fullscreen Exclusive Mode.

So, I can suggest to people that comes from others O.S. (like me from Seven) to format and reinstall, because many issues seems to be related to the upgrade procedure.

Thanks again and sorry for submitting a non-related issue.

[sebdelsol]

Oh that's awesome but a little bit complicated for what I want to do.
Could you share python code sample for enabling smooth motion?
That would help me a lot.

As far as I know, you can only access a madVR object that you've created in your process.
So you can't enable smooth motion without having the whole filterGraph (=player) running in your own program, hence the complicated part.
Maybe there's a way to do it from an external process, we should ask madshi about that :)

EDIT: You might use madVR's keyboard shortcuts instead of trying to access madVR COM interface.
Your script would send keystrokes to the madVR window to toggle smooth motion for instance.
I don't have any code sample for that though :( but I guess it's simple enough.

EDIT2 : Here (http://pastebin.com/m20Gaz5v) is a quick and dirty Python script to toggle madVR debug OSD (ctrl+J) - you could use any other shorcut.
You'll need to find your player window's handle first, which is quite easy.

EDIT3 : Oh there's another solution: You could use the SetSettings command of the madVR netProtocol.
I've never tried that, but it should work too. See net-protocol.txt in your madVR folder.

[lanzorg]

As far as I know, you can only access a madVR object that you've created in your process.
So you can't enable smooth motion without having the whole filterGraph (=player) running in your own program, hence the complicated part.
Maybe there's a way to do it from an external process, we should ask madshi about that :)

EDIT: You might use madVR's keyboard shortcuts instead of trying to access madVR COM interface.
Your script would send keystrokes to the madVR window to toggle smooth motion for instance.
I don't have any code sample for that though :( but I guess it's simple enough.

OK, I have done what I wanted using settings.bin backup files to make "presets" and UIAutomation to configure display options.

Thanks for the explications, I was completely wrong.

[chros]

OK, I have done what I wanted using settings.bin backup files to make "presets" and UIAutomation to configure display options.

Thanks for the explications, I was completely wrong.
This is my solution which relies on keyboard shortcuts and MPC-HC : http://forum.doom9.org/showthread.php?p=1735591#post1735591
(You can trigger smoothmotion via keyboard shortcut.)

[nijiko]

89.6 sometimes makes snapping image failed as the attached image.
MsgBox means "GetDIB failed" in English.

[Patrik G]

Question:
Does all these picture enhancements like chroma upscaling/error diffusion/image doubling raise the transfer rate through the HDMI cables?
is more data sent with this on compared to a regular blu ray player that doesnt have these features ?

[huhn]

there is no general answer to this.

it depends alot on settings.

Pc can output 8-12 bit RGB/YCbCr and a bd player too. if a bd player doesn't upscale it self than madVR with upscaling is kind of sending more data.

edit: the things you listed aren't really picture enchantment they are necessary steps that needs to be done at some point if scaling is needed.

[Patrik G]

there is no general answer to this.
it depends alot on settings.


i also run a 3D LUT file with madvr
dont know it that also pushes up the transfer rate?

why i asked was mainly because i could see an improvement when switching from a cheap noname HDMI cable to a 3m Supra cable to my PC.
the improvement was higher sharpness mostly visible around peoples edges.
the added sharpness was also there in motion.
movies with a alot of grain before look more clean now with less or no grain.

if i go from the above the cheaper hdmi cable was not able to transfer all of the data to the reciever.
if it would i shouldnt see any improvement right.

i also read that the Pioneer BDP LX58 blu ray player that i owned before was "pushing" the limits of what hdmi is capable off.
dont know what they mean by that.
if its transferspeeds or data sent that is maxed out.
that blu ray player also has some upscaling features on but its not close to the performance you get from madvr.

[Asmodian]

i also run a 3D LUT file with madvr
dont know it that also pushes up the transfer rate?

Only higher resolution, higher bit-depth, or higher frame rate increase the bandwidth required (assuming using the same color space). Nothing else changes the bandwidth. No setting in madVR will increase the bandwidth required over HDMI.

10-bit output would but the bit-depth actually sent to the TV is set in the GPU's drivers, not madVR.

Oh, madVR's display mode changer could change the bandwidth required, of course.

why i asked was mainly because i could see an improvement when switching from a cheap noname HDMI cable to a 3m Supra cable to my PC.
the improvement was higher sharpness mostly visible around peoples edges.
the added sharpness was also there in motion.
movies with a alot of grain before look more clean now with less or no grain.

if i go from the above the cheaper hdmi cable was not able to transfer all of the data to the reciever.
if it would i shouldnt see any improvement right.

i also read that the Pioneer BDP LX58 blu ray player that i owned before was "pushing" the limits of what hdmi is capable off.
dont know what they mean by that.
if its transferspeeds or data sent that is maxed out.
that blu ray player also has some upscaling features on but its not close to the performance you get from madvr.

I am sorry but I believe the better HDMI cable was a placebo effect. HDMI is digital and the image will be horribly corrupted or simply blank if any information is lost during the transfer.

It is possible to use a different color space with sub-sampled chroma (YCbCr 4:2:2 or YCbCr 4:2:0) which is lower bandwidth and lower quality but it will not automatically switch to this mode due to the cable's quality.

I am not sure what people were talking about with "pushing the limits of what HDMI is capable of", probably trying to sell a gold plated HDMI cable.

[Patrik G]

I am not sure what people were talking about with "pushing the limits of what HDMI is capable of", probably trying to sell a gold plated HDMI cable.

ok thanks for the info
just as i thought but im no expert here.

it was the LX-58 blu ray player that was "pushing" the limits of hdmi.
and i asume they was speaking of the picture enhancements/scaling from the blu ray player.
but its probably marketing hype.

lets see if i can catch the sharpness improvements on the camera with the different cables.
if not there then its the famous placebo effect.

[XMonarchY]

Only higher resolution, higher bit-depth, or higher frame rate increase the bandwidth required (assuming using the same color space). Nothing else changes the bandwidth. No setting in madVR will increase the bandwidth required over HDMI.

10-bit output would but the bit-depth actually sent to the TV is set in the GPU's drivers, not madVR.

Oh, madVR's display mode changer could change the bandwidth required, of course.



I am sorry but I believe the better HDMI cable was a placebo effect. HDMI is digital and the image will be horribly corrupted or simply blank if any information is lost during the transfer.

It is possible to use a different color space with sub-sampled chroma (YCbCr 4:2:2 or YCbCr 4:2:0) which is lower bandwidth and lower quality but it will not automatically switch to this mode due to the cable's quality.

I am not sure what people were talking about with "pushing the limits of what HDMI is capable of", probably trying to sell a gold plated HDMI cable.


I had to purchase a High-Speed HDMI cable to be able to use 12bit NVidia CP setting @ 60Hz on my HDMI 1.3 Samsung HDTV. With regular HDMI cable, I could use 12bit NVidia CP setting @ only 23Hz. If I tried to use 12bit setting @ 60Hz with regular HDMI cable, I would get weird artifacts all over the screen.

But I guess you are right because 12bit NVidia CP setting is not a part of madVR settings. I was just thinking if madVR is set to 10bit while a display with 12bit setting capability in NVidia CP is set to 8bit, then wouldn't madVR sort of semi-force that display to use its 12bit setting?

I had a game that required me to use High-Speed HDMI cable AND required me to set NVidia CP color depth to 12bit OR ELSE there would be many weird artifacts, the same ones I was getting when I tried to play my videos @ 12bit NVidia CP color depth and 60Hz refresh rate. I tested several regular HDMI cables to rule out cables as the cause of the issue and each time these artifacts persisted. Aside from that, my PC video and game visuals worked perfectly. The game only showed such artifacts in FullScreen (just like madVR required FSE mode to use 10bit+), but Borderless Window mode worked fine. Then, I learned about High-Speed HDMI cables and bought one. It fixed EVERYTHING - no more artifacts @ 12bit NVidia CP color depth setting @ 60Hz in neither games nor videos!

My point is that an application's request for 10/12bit color depth in Full-Screen mode @ 60hz DOES require higher HDMI bandwidth.

[jhughy2010]

I recently tried to play a 2160p home video file for the first time and madVR reported that the rendering time was over 30ms on average. I am using an HD4600 iGPU and am wondering what might be some optimal settings in madVR to lower the rendering time.

[Asmodian]

I was just thinking if madVR is set to 10bit while a display with 12bit setting capability in NVidia CP is set to 8bit, then wouldn't madVR sort of semi-force that display to use its 12bit setting?

No, in that case the GPU driver dithers (or rounds!) down to 8-bit. madVR cannot force an output bit depth.

My point is that an application's request for 10/12bit color depth in Full-Screen mode @ 60hz DOES require higher HDMI bandwidth.

Yes, as I said: resolution, bit-depth, and frame rate all influence the bandwidth required. You do need an HDMI cable that can handle the signal sent from the GPU to the TV. No setting in madVR will affect these (besides the resolution and refresh rate changer).

However, as you have noticed, the result when using a cable that cannot handle the bandwidth is horrible artifacts, flickering, or even no image. It never results in slightly softer edges around people; that type of signal degradation is only a concern with analogue transmission. ;)

edit:
In a digital format, if electrical noise causes a pixel to change color, it will change color by a small amount with the same probability as a large amount. (11001000,00000000,01100101) to (11001100,00000000,01100101) or (11001000,10000000,01100101), the second is a very different color but both are a single bit flip from a 0 to a 1.

[Magik Mark]

FYI, New Nvidia driver v358.50 Supports OpenGL 2015 ARB extensions and OpenGL ES 3.2

Does it have any significance to madvr?

[huhn]

madVR doesn't use OpenGL

[bouwew]

I found a small issue with MadVR 0.89.6, using MPC HC as a player. It's also present in 0.89.5.
After a video is started, running FSE, I right click to change some settings in Madvr, LAV, etc. When I exit after the changes made, the screen returns to FSE and the playback is stuttery and the audio is a little out-of-sync. This effect can be "removed" by switching away and back to full screen.
Anyone else experience this?

Windows 10 x64, AMD 15.9.1, running latest MPC-HC & LAV filters nightly.

[galneon]

I found a small issue with MadVR 0.89.6, using MPC HC as a player. It's also present in 0.89.5.
After a video is started, running FSE, I right click to change some settings in Madvr, LAV, etc. When I exit after the changes made, the screen returns to FSE and the playback is stuttery and the audio is a little out-of-sync. This effect can be "removed" by switching away and back to full screen.
Anyone else experience this?

Windows 10 x64, AMD 15.9.1, running latest MPC-HC & LAV filters nightly.

I've gotten similar stuttering for several versions now when switching from windowed to FSE (settings menus not required). It doesn't always happen. The quickest solution for me is simply right-clicking and then closing the context menu. The context menu switches you to windowed full screen and when it switches back to FSE a few seconds later the problem is remedied every time. If I double-click two or four times, I'm back in FSE with stuttering.

I've had a similar problem with MPCHC hanging on every exit but the crash reporter window doesn't come up.

I fixed it by cleaning out all codecs and reinstalling, upon which time I switched to the x64 version of MPCHC.

Unfortunately I had to redo all the settings but there was no way around it.

I think I've waited long enough to make the 64-bit transition myself now that all my favorite filters are available for 64-bit MPC-HC. Thanks, I think that will probably take care of it.

[edigee]

Acc. to reports on other forums the latest NVIDIA driver(358.50) fixed Nnedi3 issue in W10.
Also it finally fixes custom resolutions .

[aufkrawall]

Acc. to reports on other forums the latest NVIDIA driver(358.50) fixed Nnedi3 issue in W10.

This was fixed my madshi by using another interop extension, the old one is still missing.

[madshi]

Were you expecting an answer? I see you haven't added it for the last three versions, is it difficult? Personally runtimeInMinutes is fine, I can't answer the your first point.
Yes, I was waiting for a "yes, that would work for me" or something. I'll put "runtimeInMinutes" on my to do list. Can't give you an ETA, though.

Still no fix for stutters after pause & resume?
I've already replied to your previous report about that, but you didn't follow up on my reply.

There's power state issue by using madVR, when I switched the image doubler from NNEDI3 64 neurons to super-xbr, there's a lot of dropped frames, and the present queue cannot be filled.

Using less gpu-intensive algorithms can decrease the speed of the fan, and hence the noise. How would I do to let the GPU chose the power-state correctly?
madVR currently does not control the GPU power state at all. It's the duty of the GPU driver to not clock down if that would cause issues. Seems your GPU driver doesn't do that correctly. That's not madVR's fault.

hi guys. i am using madvr version 0.88.16.
i am having a problem with a single avi file.
i use my display mode swith like this:
1080p60, 1080p59, 1080p50, 1080p30, 1080p29, 1080p25, 1080p24, 1080p23
my avi file is at 29,970fps but when i go fullscreen at madvr it changes the screen refresh rate to 24p, so the video is unwatchable.
when i delete all the lines and leave only 1080p29 it go to correct screen refresh rate. but if i add 1080p29, 1080p23, it prefer 1080p23.
anyone know how to solve this?
What does the debug OSD (Ctrl+J) say about the movie xxx fps? The full text line, please.

Subtitles glitch out now. When a file is first launched and full-screen'd, subs don't render until you right click the video frame or do something else to kick it out of full screen exclusive. From that point on they render fine.

Problem is fixed upon reverting to 0.89.5

Edit: This is on xySubfilter/LAVsplitter/MPCHC btw. Running in D3D11 render path, using FSE, with 64 bit build.
Confirmed. Will try to fix this for the next build.

I have a little request regarding toggle debanding keyboard shortcut behaviour. Let me explain a bit first. I usually use either a high debanding or no debanding during several files in a series. If I'd switch through GUI, it's all plain and simple - the radiobutton is always at "high" position and I can just toggle only the checkbox. Obviously I want to speed up things by using keyboard shortcuts, but the different story begins here. For example, I start from debanding at "high" but off and want to turn it on. So I press Ctlr-Alt-Shift-A twice (OSD shows off->high) and then hit F2 to save. Everything goes as expected. Afterwards, I want to switch it back off and repeat the same sequence, OSD showing high->off. But now it turns out that not only the checkbox was toggled - the radiobutton was also set to "low" value, so next time I'd need to press Ctlr-Alt-Shift-A four times, going all the way through off->low->medium->high. I humbly ask you to change this behaviour so that the radiobutton won't get reset from its initial value if the final result of keypresses is "off" at the moment of releasing all keys in the shortcut combination. It's not that I'm too bothered with two extra keypresses - I just want to get used to the same keypress sequence for toggling debanding on and off and pay no attention to its initial state or OSD.
Ok, I'll see what I can do, but it might take some time...

Does anyone know how to use madvr developers interface with python or powershell?
Is IMadVRSettings available as COM object?
As far as I know, you can only access a madVR object that you've created in your process.
So you can't enable smooth motion without having the whole filterGraph (=player) running in your own program, hence the complicated part.
Maybe there's a way to do it from an external process, we should ask madshi about that :)

EDIT3 : Oh there's another solution: You could use the SetSettings command of the madVR netProtocol.
I've never tried that, but it should work too. See net-protocol.txt in your madVR folder.
madHcCtrl.exe (which is the exe responsible for drawing the madVR tray icon, showing the madVR settings editor etc) does all this by using madHcNet32.dll and mvrSettings32.dll. I've not made the interfaces for those dlls available yet because I didn't think they would be useful. But I suppose it wouldn't harm making them available. It would allow you to remotely access settings, without having to reimplement the whole network protocol yourself again.

The above applies if you want to actually affect an already running instance of a media player / madVR. If you just want to change the settings on harddisk, and if you don't care if already running madVR instances are updated, then things are much easier: Just instantiate a madVR instance yourself and use the interfaces in "mvrInterfaces.h" to modify the settings.

But in the meantime during I was thinking about the idea madshi came across with these nice settings with the black borders...
So the second solution would be more of a software based instead of the hardware based photodiode.
But that would require to gain the necessary information out of madvr. That would mean something like target rectangle out of an API. I can do a little bit of programming in C# (not very advanced, but I can do stuff like GUIs, automats,...) so no problem to implement the controlling of my masking setup eg by RS232 to get the information out of the PC. But I don´t know if A) I can get the necessary information out of madvr and B) if yes how.
If this is too complicated I would try the version with the photodiodes. The software based solution seems more reliable to me, so I would prefer this one ;)
I think the software solution should be much more reliable. I've added the "command line to execute" feature to profiles for the exact reason that people might want to automate anamorphic lenses or screen masking or lens memories or stuff. So the plan was that every time a profile is switched, madVR calls some command line, maybe passing some parameters. However, I'm not sure if that's what you need? It would allow you to e.g. switch between CinemaScope and 1.85:1 movies. But maybe you want a "perfect" solution which tells you exactly (with 1 pixel precision) where the black bars are? That's somewhat more complicated because currently I've no feature in madVR that would transport that kind of information to you. Shouldn't be hard to add, but the big question is how to do it in a useful and intuitive way, so that other users could reuse the same feature...

Has anyone else experienced an MPC-HC crash on every exit for the last several minor versions of madVR? Only way to close the crash reporter window is killing the process in task manager. Windows 7 64-bit, 32-bit MPC-HC/madVR, GTX 970. I can give more details if there isn't a known crash that behaves in this way. I haven't experimented with settings lately to find if there's a particular cause because I've assumed the crash would go away after the next update, but many updates later I'm starting to wonder if it isn't a known issue.
No such problems here. Do you get any kind of crash information at all? If it's MPC-HC that crashes then there's probably not much I can do to help and you might have to talk to the MPC-HC devs. Might still be my fault, but without a proper crash report, *and* without being able to reproduce it at the same time, I simply cannot do anything.

How should we tune "Zoom control" options to follow rule "If Zoom do not lose a single pixel of video only black bars"?
That's madVR's default behaviour, if you activate "automatically detect hard coded black bars" (without any sub options) and set your media player to "touch window from inside".

It seems madVR is causing MPC-hc to crash if NVidia 3D vision is enabled when going into fullscreen? If I set MPC-HC to use EVR then it does not crash. Is there any way I can make a meaning full log for you to track this one given madvr does not seem to crash?
I don't think it's likely to be my fault. I don't even know if NVidia 3D vision is enabled or not. If the crash is in MPC-HC then you could talk to the MPC-HC devs. Maybe they can provide some more information about where the crash occurs and why.

Hope you can find the time to add support for the MP4 rotation flag. :)
Is it a flag in the container or in the video bitstream? If it's in the container then madVR has no access to that at all.

89.6 sometimes makes snapping image failed as the attached image.
MsgBox means "GetDIB failed" in English.
Can you try to find out how "sometimes" is defined exactly? Is it randomly, sometimes works, sometimes not, with the same video file and settings? Or does it depends on the settings and the video file etc?

I recently tried to play a 2160p home video file for the first time and madVR reported that the rendering time was over 30ms on average. I am using an HD4600 iGPU and am wondering what might be some optimal settings in madVR to lower the rendering time.
30ms is ok if the video file is 24fps. If you want faster rendering speeds, probably the settings to look at are "chroma upscaling" and "image downscaling".

I found a small issue with MadVR 0.89.6, using MPC HC as a player. It's also present in 0.89.5.
After a video is started, running FSE, I right click to change some settings in Madvr, LAV, etc. When I exit after the changes made, the screen returns to FSE and the playback is stuttery and the audio is a little out-of-sync. This effect can be "removed" by switching away and back to full screen.
Anyone else experience this?
I've gotten similar stuttering for several versions now when switching from windowed to FSE (settings menus not required). It doesn't always happen. The quickest solution for me is simply right-clicking and then closing the context menu. The context menu switches you to windowed full screen and when it switches back to FSE a few seconds later the problem is remedied every time. If I double-click two or four times, I'm back in FSE with stuttering.
Don't have that problem here. You can try capturing it in a debug log. If you have that stuttering situation, let it run for 20 seconds with stuttering, then immediately close the media player. Don't switch back to windowed mode or try to fix the stuttering. Just let it stutter for 20 seconds, then close the media player as quickly with as little user intervention as possible. Probably Alt+F4 should do the trick.

[har3inger]

There's power state issue by using madVR, when I switched the image doubler from NNEDI3 64 neurons to super-xbr, there's a lot of dropped frames, and the present queue cannot be filled.

Using less gpu-intensive algorithms can decrease the speed of the fan, and hence the noise. How would I do to let the GPU chose the power-state correctly?

Setting application-specific power settings has to be done for the player that madvr is being used inside. For example, if you use MPCHC, you have to force your GPU to treat mpchc.exe in such and such a way. In some ATI drivers (and possibly some nVidia ones), this exe is locked to power saving. You may need to rename the executable to something else (like mpchcFIX.exe) to get power settings to work. Beyond that, the specifics of how to do that will depend on your unique setup.

[aufkrawall]

Situation now is that interop issue with AMD is gone with the new extension used for NNEDI3, but the computing itself is still slower with drivers after 13.12?

[QBhd]

Situation now is that interop issue with AMD is gone with the new extension used for NNEDI3, but the computing itself is still slower with drivers after 13.12?

Yes, 13.12 is still far and away the fastest drivers for madVR. But as a guy who does overkill, I switched to the latest drivers and still was able to boost a few settings.

QB

[nijiko]

Can you try to find out how "sometimes" is defined exactly? Is it randomly, sometimes works, sometimes not, with the same video file and settings? Or does it depends on the settings and the video file etc?

Randomly.
I also don't know when it will happen.
40%.
With everyone video files.

I think it is maybe failed with videocard memory, because I set madVR to use DXVA.
And this happens on WinXP with DXVA1.

[bouwew]

Don't have that problem here. You can try capturing it in a debug log. If you have that stuttering situation, let it run for 20 seconds with stuttering, then immediately close the media player. Don't switch back to windowed mode or try to fix the stuttering. Just let it stutter for 20 seconds, then close the media player as quickly with as little user intervention as possible. Probably Alt+F4 should do the trick.

Debug log here: http://hugefiles.net/3a885fpkawyt

[leeperry]

I switched to the latest drivers and still was able to boost a few settings.
As compared to? Of course latest mVR builds come with faster NNEDI3 to boot, even with AMD IIRC.

Situation now is that interop issue with AMD is gone with the new extension used for NNEDI3, but the computing itself is still slower with drivers after 13.12?
But no option to output 10bit with 13.12 AFAIK, you can apparently run a reghack to disable 8bit dithering but latest revs do come with an official option to output >8bit AFAIK.

That's also why I'm not too worried about 10bit with mVR coz 13.12 work like a charm for me and PQ is stellar in 8bit@ED2. OCD can find its limits if you decide so ><

madshi can you share your future plans for madVR?
Moar better PQ, no less.

Extremely impressed by the PQ of DVD's going through sxbr25 + chroma & luma SR, who knew you could squeeze out so many details out of SD huh.

I've had a good look at what SR does to 24p motion blur and on scenes with wind going through tree leaves it basically ensures that all frames get equally sharp, so there is no blurry interframes anymore and you do get 24 sharp frames for a change.....surely it kinda kills the "dreamy" feeling you're supposed to get from vanilla 24p and it does look more artificial but it's by very far the best alternative I've seen between 60p interpolation and raw 24p, not going back to stutterland TYVM :)

We used to enjoy some slight LSF on HCFR ages ago due to this very same effect on motion blur but it looked too unsharpish to be useful to anything and Avisynth in ffdshow is all just a big lossy 8bit pipeline.

Of course for native 1080p footage I gotta zoom one notch in order to have SR kicking in, hopefully madshi will implement SuperSampling someday.

I really find sxbr a nice in-between J3AR & NNEDI3, being a sharper version of the former and a less artificial/digital looking alternative to the latter.

Of course soon or later madshi will come up with another groundbreaking new feature that will render it all obsolete, the same way J3AR & then NNEDI3 used to be on the very top of my list not so long ago ^^

[amggotleir]

allright guys, looks like there has been some big updates, can some1 tell me the latest settings for the best possible image quality?

[aufkrawall]

Yes, 13.12 is still far and away the fastest drivers for madVR. But as a guy who does overkill, I switched to the latest drivers and still was able to boost a few settings.

Oops, I confused interop and copyback.
I mean, now there definitely is no more copyback with AMD?

[AngelGraves13]

allright guys, looks like there has been some big updates, can some1 tell me the latest settings for the best possible image quality?

Jinc for Chroma and Image Upscaling with AR enabled for both.
Downscaling is anyone's guess... I just look at the chart on the right and try to get zero ringing and very little aliasing. Spline has a bit of ringing, so I use Lanczoz.

Super-XBR is also nice for Chroma, but tends to be too sharp sometimes so you'll have to play with the sharpness. 50 seems ok though, but not any better than Jinc.

[aufkrawall]

Jinc for Chroma and Image Upscaling with AR enabled.

AR should also be enabled for chroma with jinc.
I stumbled over some nasty news animation with red arrows and you can clearly see every upscaling algorithm fail which is not utterly soft or NNEDI3.
But Jinc AR looks good in many situations and doesn't eat up too much performance.


Downscaling is anyone's guess... I just look at the chart on the right and try to get zero ringing and very little aliasing. Spline has a bit of ringing, so I use Lanczoz.

Catmull-Rom AR is often preferred for downscaling, as it's neither overly sharp nor soft.
It's currently not clear if linear light should be used or not. There are examples where it introduces aliasing.
I personally don't use it, as I rather like the result slightly darker instead of brighter, if I have to choose between both.

[Asmodian]

It's currently not clear if linear light should be used or not. There are examples where it introduces aliasing.
I personally don't use it, as I rather like the result slightly darker instead of brighter, if I have to choose between both.

When downscaling content from its native resolution I always use linear light because small spots of light (stars) fade or disappear without it.

However, when downscaling previously up-sampled material there are never spots of light small enough to be eliminated by downscaling in gamma light.

[AngelGraves13]

AR should also be enabled for chroma with jinc.
I stumbled over some nasty news animation with red arrows and you can clearly see every upscaling algorithm fail which is not utterly soft or NNEDI3.
But Jinc AR looks good in many situations and doesn't eat up too much performance.


Catmull-Rom AR is often preferred for downscaling, as it's neither overly sharp nor soft.
It's currently not clear if linear light should be used or not. There are examples where it introduces aliasing.
I personally don't use it, as I rather like the result slightly darker instead of brighter, if I have to choose between both.

I use AR for both chroma and upscaling. Should probably word it better.

Catmull-Rom has too much aliasing if I look at the chart in madVR, but it's really hard to notice a difference between any of them. You'll be upscaling more likely than downscaling.

[huhn]

use your eyes not the chart!

[FIX94]

I personally use Jinc AR for both chroma and image upscaling, it just gets the job done right, for downscaling I use Spline36 AR cause no matter what people have against Spline it sure does a good job at downscaling and even acceptable upscaling to my eyes :P

[ikarad]

I use AR for both chroma and upscaling. Should probably word it better.

Catmull-Rom has too much aliasing if I look at the chart in madVR, but it's really hard to notice a difference between any of them. You'll be upscaling more likely than downscaling.

Where is the chart?

[Asmodian]

Where is the chart?

The "image upscaling" and "image downscaling" pages in the "scaling algorithms" section of madVR's settings.

[Magik Mark]

Madshi,

Is there a way for madvr to take advantage of SLI? I'm planning to install one and hope would help in SVP + madvr using the most advance upscaling algorithm offered

[aufkrawall]

We had this a few times:
No, this is not possible nor planned.

[madshi]

Randomly.
I also don't know when it will happen.
40%.
With everyone video files.
In that case I'm not sure where it comes from. Is this a new problem? Or was it always this way?

Debug log here: http://hugefiles.net/3a885fpkawyt
Thanks, will have a look at it later.

I mean, now there definitely is no more copyback with AMD?
Hard to say, haven't intensively tested it. But so far several users which had severe performance problems with AMD (probably due to copyback) have reported much better performance with the new madVR builds. However, some users also don't noticed any performance gain. So it's hard to give a definite answer. My current impression is that there's probably no copyback, anymore. But this doesn't give a general performance boost, it just removes a bottleneck *some* users had problems with.

Is there a way for madvr to take advantage of SLI? I'm planning to install one and hope would help in SVP + madvr using the most advance upscaling algorithm offered
madVR itself cannot use SLI. *However*, I think SVP lets you choose which GPU to use for its motion estimation. So with an SLI system you might be able to move the SVP workload onto the 2nd GPU while using the main GPU for madVR. I haven't tested this, though, so I can't say if it works and whether you get a real performance benefit this way or not.

[madshi]

madVR v0.89.7 released

http://madshi.net/madVR.zip

* added support for new XySubFilter build -> full control over positioning
* added IMadVRInfo::GetRect("seekbarRect")
* fixed: XySubFilter subtitles sometimes only showed up after resizing video
* fixed: #318: debanding sometimes caused black screen
* fixed: #344: NNEDI3 doubling + smooth motion sometimes caused black screen
Originally I was planning to work on SuperRes and other stuff. But instead I've decided to concentrate on XySubFilter this weekend. I've hacked a minor addition into XySubFilter, to give madVR full control over where subtitles are drawn. Which means that with the new madVR build, in combination with the new XySubFilter build, subtitles are now moved into the black bars or into the active video area, depending on the situation.

The latest XySubFilter build is now available in the first post of the official XySubFilter thread:

http://forum.doom9.org/showthread.php?t=168282