madVR - high quality video renderer (GPU assisted)

[madshi]

Yeah, maybe. Don't have fun/time to add that right now, though.

FWIW, I've enough results from ATI users now. But some more NVidia + Intel results would be nice.

[renq]

GTX460 results sent ;)
FW 267.05, should it matter

[Razoola]

Yeah, maybe. Don't have fun/time to add that right now, though.

FWIW, I've enough results from ATI users now. But some more NVidia + Intel results would be nice.

I also have an issue where I have 2 gpus, trying to run the test program on the secondary gpu is a pain given its display is in a different room from the PC. I have managed to do it but it was a real haedache. If you do update it adding driver details adding a switch to choose the monitor to run the test on could be really helpfull.

This result suffers the problem you mention about colors although I do already have the suggested fixes you mention in place. I want to point out that this gpu is output via HMDI. Will send the result now.

[DottorLeo]

Ati 5770 result send! :D

[nevcairiel]

If you have an NVidia GPU and if in the "D3D9 Surface StretchRect" section the test results say "(0-255)" that means that your NVidia GPU outputs video levels, by forcefully stretching all data behind madVR's back. This can result in banding artifacts. The fix for this problem is to create custom resolutions. You can double check whether the fix works by running madNV12Test again. If it reads "(16-235)" in the "D3D9 Surface StretchRect" section, that means you're getting proper output levels for best madVR quality.

I have custom resolutions on my HTPC, yet it still says 0-255 in the output of your tool. I have not seen any evidence of banding, however. You wouldn't happen to have an easy to use sample file? :d

[nevcairiel]

I have custom resolutions on my HTPC, yet it still says 0-255 in the output of your tool. I have not seen any evidence of banding, however. You wouldn't happen to have an easy to use sample file? :d

I previously had the Dynamic Range configured to 0-255, because my TV was supposed to get PC Level output. If i set that to Video Player controlled, i get the proper values in your tool.

I suppose thats how its meant to be configured to let madVR expand it to PC levels, right?

[Razoola]

I have custom resolutions on my HTPC, yet it still says 0-255 in the output of your tool. I have not seen any evidence of banding, however. You wouldn't happen to have an easy to use sample file? :d

I have been playing with this also since madshi has brought it up and I have not seen any banding either. This issue I have however is if I do change the nvidia HMDI setting to output desktop levels (it was set to full screen videos) then I get a washed out picture on my plasma. I think this setting is only available when using HDMI out from the gfx card and is found at the bottom of the 'adjust desktop color settings' page in the nvidia control panel.

Having this set to RGB and full screen video gives me the best full contrast picture.

[madshi]

I have custom resolutions on my HTPC, yet it still says 0-255 in the output of your tool. I have not seen any evidence of banding, however.
Make sure you set "Adjust video color settings - Advanced - Dynamic Range" to "Video Player Settings" in your NVidia control panel. Otherwise the "0-255" vs. "16-235" output of my tool doesn't say much about whether you'll get banding or not.

You wouldn't happen to have an easy to use sample file? :d
Sure! Scale the "smallramp.ytp" test pattern up to full screen resolution. That should produce a very smooth grayscale. (see madTestPatternSource folder).

[namaiki]

There will/may be banding visible in EVR and any of the other system renderers, (VMR and Overlay) if you set your graphics card to expand from TV to PC levels without dither. (nVidia 9600M GT and Intel 4500MHD which I tested do not dither) (not saying you didn't already know this)

[nevcairiel]

Sure! Scale the "smallramp.ytp" test pattern up to full screen resolution. That should produce a very smooth grayscale. (see madTestPatternSource folder).
That does look pretty smooth to me.

[madshi]

I have been playing with this also since madshi has brought it up and I have not seen any banding either. This issue I have however is if I do change the nvidia HMDI setting to output desktop levels (it was set to full screen videos) then I get a washed out picture on my plasma. I think this setting is only available when using HDMI out from the gfx card and is found at the bottom of the 'adjust desktop color settings' page in the nvidia control panel.
Weird, never seen this setting before. Not sure which practical effect it has. From the description it seems it's only an information that is sent to the display. Either your plasma reacts to it in a weird way, or NVidia does more than just sending an information.

[madshi]

I previously had the Dynamic Range configured to 0-255, because my TV was supposed to get PC Level output. If i set that to Video Player controlled, i get the proper values in your tool.

I suppose thats how its meant to be configured to let madVR expand it to PC levels, right?
Actually this setting should have no effect on madVR at all. It does affect the madNV12Test tool, though. Having it set to "Video Player controlled" allows the madNV12Test tool to find out whether your GPU outputs PC vs. video levels (this only applies to NVidia GPUs). Since you say you get the proper values now, everything should be fine. You can set the settings back to 0-255, if you want. Shouldn't bother madVR. It might have an effect on VMR/EVR, though.

[nevcairiel]

Have you checked "Adjust video color settings - Advanced - Dynamic Range"? What is this set to on your PC?

I set it to video player controlled now (which also changed the output of your tool to 16-235).

Fwiw, i do not see any difference between video player controlled, or setting PC/Video there

[madshi]

That's as its supposed to be, see my comment above yours.

[Razoola]

For my GT240 all the suggested changes you suggest do not change the way you test program reports, I still get 0-255.

This is 2nd gpu outputting to a secondary display. Like I said however I have to make the gt240 the primary display and gtx295 secondary to get the program to report details for it. I have tried all the suggested settings you suggest and it just won't go from 0-255 to 16-235.

[madshi]

Hmmmm. You have created and activated a custom resolution? Is the "smallramp.ytp" test pattern perfectly smooth if you upscale it to screen resolution?

[Razoola]

If I create my own test custom resolution it works but If I use the nvidia preset custom resolutions it does not.. Here lies the problem, I cannot create the custom resolutions I need because it messes up all the other standard resolutions in such a way that I have to clean reinstall the driver. This is an issue I think with trying to create custom resolutions on secondary displays in the nvidia control panel, it sucks. This bug has been there for ages and nvidia have not fixed it even though they are aware of it.

[yesgrey]

This is an issue I think with trying to create custom resolutions on secondary displays in the nvidia control panel, it sucks. This bug has been there for ages and nvidia have not fixed it even though they are aware of it.
I know. Did you try to create the custom resolutions on the secondary display by activating it and disabling the primary display, so it can be the only active? Whenever I do this I can create my custom resolutions on both displays without any problem. The secret is creating the custom resolutions only with one of the displays active, the one which you want to create the resolutions to.;)

[Razoola]

I know. Did you try to create the custom resolutions on the secondary display by activating it and disabling the primary display, so it can be the only active? Whenever I do this I can create my custom resolutions on both displays without any problem. The secret is creating the custom resolutions only with one of the displays active, the one which you want to create the resolutions to.;)

I'll give it a go and report back.

Tried, as soon as I enter the first resolution all other resolutions are limited to 16bit.

[mark0077]

For those of you that didn't know, for whoever has a HDTV connected with nvidia drivers, and the entire desktop is being output in 16-235 instead of full range, the best workaround is to add this to the nv_disp.inf before installing the drivers. Makes the drivers assume your using a fullrange TV instead of limited range. I need to use this each time I do a clean install of nvidias drivers.

Add

HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

to each section of your inf that begins with

[nv_miscBase_addreg_

eg

[nv_miscBase_addreg__01]
HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

======================

madshi, a general question. You might remember I recently reported about having less frame drops in some particular files, when using high performance mode of my nvidia card, rather than adaptive performance. Well it turns out I'm having issues with many of these files in another way which I havn't been able to figure out, with getting much more frame drops than others even when not using any avisynth scripts, with cpu usage less than 15%. The evr-cp renderer in mpc actually showed me with its graph, that the presentation of video isn't smooth with these files compared to blu-rays / other files on my machine. See below the green line isn't straight when playing these files.

http://img832.imageshack.us/img832/6956/waveso.th.jpg (http://img832.imageshack.us/i/waveso.jpg/)

I still havn't figure it out but am wondering does madVR do any type of timestamp smoothing or is it something you think might be useful. So far all I can think of that causing this effect with these files is something to do with timestamps. Some are mkv, some are mp4 and I havn't tried too many more files to see how many it effects. (PS: The evr-cp graphs / or the unsmooth presentation would have come in handy in madVR in this case :) ). Even when there are 0 frame drops its easy to see from the graph something isn't right..

[namaiki]

HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

This should be useful for me since I've been using VGA because of that issue.

Would you happen to know how I might be able to enable a 24Hz (or multiple of 24Hz) mode on my TV? At the moment, I've only got bloody 25Hz and 60Hz.

[mark0077]

This should be useful for me since I've been using VGA because of that issue.

Would you happen to know how I might be able to enable a 24Hz (or multiple of 24Hz) mode on my TV? At the moment, I've only got bloody 25Hz and 60Hz.

Yes, exactly, the other workaround I had was some edid override which limited the hz that were possible.

The workaround I gave doesn't have any negative effects... pity there isn't a gui based option...

[Razoola]

For those of you that didn't know, for whoever has a HDTV connected with nvidia drivers, and the entire desktop is being output in 16-235 instead of full range, the best workaround is to add this to the nv_disp.inf before installing the drivers. Makes the drivers assume your using a fullrange TV instead of limited range. I need to use this each time I do a clean install of nvidias drivers.

Add

HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

to each section of your inf that begins with

[nv_miscBase_addreg_

eg

[nv_miscBase_addreg__01]
HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

======================

madshi, a general question. You might remember I recently reported about having less frame drops in some particular files, when using high performance mode of my nvidia card, rather than adaptive performance. Well it turns out I'm having issues with many of these files in another way which I havn't been able to figure out, with getting much more frame drops than others even when not using any avisynth scripts, with cpu usage less than 15%. The evr-cp renderer in mpc actually showed me with its graph, that the presentation of video isn't smooth with these files compared to blu-rays / other files on my machine. See below the green line isn't straight when playing these files.

http://img832.imageshack.us/img832/6956/waveso.th.jpg (http://img832.imageshack.us/i/waveso.jpg/)

I still havn't figure it out but am wondering does madVR do any type of timestamp smoothing or is it something you think might be useful. So far all I can think of that causing this effect with these files is something to do with timestamps. Some are mkv, some are mp4 and I havn't tried too many more files to see how many it effects. (PS: The evr-cp graphs / or the unsmooth presentation would have come in handy in madVR in this case :) ). Even when there are 0 frame drops its easy to see from the graph something isn't right..

Are you sure you can't simply use the setting I mentioned earlier at the bottom of the 'adjust desktop color settings' page in the nvidia control panel?

There are 2 settings that show if your outputting via HDMI, you can choose either RGB or YCbCr444 for digital color format. Also you can choose either auto select, desktop programs or full screen video for content type reported to the display.

When ever I update the driver I have to do these settings again to get full range back.

[madshi]

For those of you that didn't know, for whoever has a HDTV connected with nvidia drivers, and the entire desktop is being output in 16-235 instead of full range, the best workaround is to add this to the nv_disp.inf before installing the drivers. Makes the drivers assume your using a fullrange TV instead of limited range. I need to use this each time I do a clean install of nvidias drivers.

Add

HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1

to each section of your inf that begins with

[nv_miscBase_addreg_

eg

[nv_miscBase_addreg__01]
HKR,,SetDefaultFullRGBRangeOnHDMI,%REG_DWORD%,1
Cool trick! But a google search tells me this only works in Vista / Windows 7, but not in XP. So for XP creating custom resolutions is still the best workaround.

madshi, a general question. You might remember I recently reported about having less frame drops in some particular files, when using high performance mode of my nvidia card, rather than adaptive performance. Well it turns out I'm having issues with many of these files in another way which I havn't been able to figure out, with getting much more frame drops than others even when not using any avisynth scripts, with cpu usage less than 15%. The evr-cp renderer in mpc actually showed me with its graph, that the presentation of video isn't smooth with these files compared to blu-rays / other files on my machine. See below the green line isn't straight when playing these files.

I still havn't figure it out but am wondering does madVR do any type of timestamp smoothing or is it something you think might be useful. So far all I can think of that causing this effect with these files is something to do with timestamps. Some are mkv, some are mp4 and I havn't tried too many more files to see how many it effects. (PS: The evr-cp graphs / or the unsmooth presentation would have come in handy in madVR in this case :) ). Even when there are 0 frame drops its easy to see from the graph something isn't right..
I'm not really sure what you're asking here. The problem seems to occur with EVR just as much as with madVR? So I'd guess it's probably either a problem with the movie files, or with your splitter/decoder setup. What kind of files are these? MKV? Broadcast (ts)? Blu-Ray (m2ts)? If it's MKV try remuxing with eac3to. Maybe that helps?

[Razoola]

Cool trick! But a google search tells me this only works in Vista / Windows 7, but not in XP. So for XP creating custom resolutions is still the best workaround.

My gtx 295 does not have custom resolutions added but it returned the correct values with your test program. It uses DVI however and not HDMI.

For now I have given up trying to get correct setting with the secondary GT240 as I cannot set up the custom resolutions I need on it using the nvidia control panel.

[mark0077]

Hi madshi, I suppose what I'm asking is

1) Does madVR do any sort of checking for uneven timestamps, and is it able to smooth them out?
2) If not, do you think its something the renderer should / or could do in future?
3) I'm going to go through alot of my content over the next few days to see which are effected, and will use evr-cp to check this, unless madVR has a way of indicating the issue I see using evr-cp's graphs, with some of its CTRL-J stats? Not a big issue I can use evr-cp if madVR hasn't a way of showing such problems up.

Here is one clip which I downloaded recently, its Lada Gaga's new music video, 1080p. I won't post a direct link but the filename is "Lady Gaga - Born This Way(1080p_H.264-AAC) VEVO.mp4" - 171 MB (179,834,004 bytes)

I don't know the source but if you wanted you could try it and see the evr-cp graphs, and see the framedrops with madVR and other renderers of course. I can only assume thousands of people will have similar files, and I would love a way to workaround the problems I'm having with them without having to remux them all :) Let me know if you have any thoughts. If madVR could workaround such problems, it would be another step ahead of evr-cp but I'm unsure if its appropriate for the renderer to do such a thing.

In the meantime I'll try remuxing one of the mkv's I have the problem with, with eac3to to see if it helps.

[madshi]

1) Does madVR do any sort of checking for uneven timestamps, and is it able to smooth them out?
2) If not, do you think its something the renderer should / or could do in future?
3) I'm going to go through alot of my content over the next few days to see which are effected, and will use evr-cp to check this, unless madVR has a way of indicating the issue I see using evr-cp's graphs, with some of its CTRL-J stats? Not a big issue I can use evr-cp if madVR hasn't a way of showing such problems up.
1) madVR swaps swapped timestamps back. But if the timestamps are not swapped, just "uneven", madVR does not do anything about it. However, if you have a 1:1 refresh rate map (24fps content with 24Hz refresh rate), uneven timestamps should automatically get smoothed out. That doesn't work if you have a higher refresh rate than frame rate, though.
2) Maybe.
3) I don't know what the issue is so I don't know if madVR shows it.

[Mr Alpha]

What is madVR doing that makes pagefile writes take off? Just got and SSD and was wondering if I'm making enough writes to the pagefile to make it worth moving it off the SSD. So I fired up Process Monitor, and made a long capture.

During normal desktop usage the pagefile writes were areound 5MB/h, but when watching something with MPC-HC and madVR the pagefile writes increase to over 3GB/h. This does not happen with EVR.

3GB per hour is quite a lot of writes. What is madVR doing?

[madshi]

The pagefile is managed by the OS. If madVR increases pagefile usage then probably your PC doesn't have enough RAM in it. madVR keeps 8 decoded frames in system RAM. I don't think EVR does that. Sorry, nothing I can do about it. The max I could do is to offer an option to decrease the queue size. I have that planned for a future build, but it may take some time until I get there.

[mark0077]

1) madVR swaps swapped timestamps back. But if the timestamps are not swapped, just "uneven", madVR does not do anything about it. However, if you have a 1:1 refresh rate map (24fps content with 24Hz refresh rate), uneven timestamps should automatically get smoothed out. That doesn't work if you have a higher refresh rate than frame rate, though.
2) Maybe.
3) I don't know what the issue is so I don't know if madVR shows it.
Thanks madshi. I'll do more tests with these files tomorrow but if it wouldn't be too much work on your part I would really appreciate the timestamp smoothing ability. I can report back If this is ever implemented and if it resolves these frame drops with these problem files I think it could improve htpc playback for those with these seemingly imperfect / badly muxed files which could be a very good thing :)

Razoola. That nvidia cp option never has any effect on my pc. I'm not sure what it does but for me I need that registry tweak to get full range with samsung series 9 - gtx295 - win 7 64 bit

[Hypernova]

What is madVR doing that makes pagefile writes take off? Just got and SSD and was wondering if I'm making enough writes to the pagefile to make it worth moving it off the SSD. So I fired up Process Monitor, and made a long capture.

During normal desktop usage the pagefile writes were areound 5MB/h, but when watching something with MPC-HC and madVR the pagefile writes increase to over 3GB/h. This does not happen with EVR.

3GB per hour is quite a lot of writes. What is madVR doing?

Could you tell me briefly how to check this? I have an SSD as well but I don't think madVR is doing that on my setup.

[bjd]

D3D9 Surface speed test:
NV12: upload 458 fps, download 27 fps, trick download failed
YV12: upload 278 fps, download 26 fps, trick download failed


Do the figures above indicate NV12 gives better/faster performance than YV12 ?

[madshi]

If you upload to NV12/YV12 surfaces then yes, it appears so. But that's not how madVR works at the moment. It doesn't matter to madVR whether you feed madVR with NV12 or YV12. Zero difference in performance.

[bjd]

@Madhsi - thanks for that, it is quite interesting.

On another note regarding my delayed frames on XP, the card is maxed at 100% performance and was still dropping/delaying frames after running a while, however i have managed to stop this by configuring the NVidia 3D settings to "Let the 3D application decide".

It would seem the more things you leave to the card driver to decide, the more things go wrong.

[madshi]

Maybe if you force anisotropic filtering or antialiasing that could also affect madVR. It could cost performance, and it could also hurt image quality.

[bjd]

I don't tend to force anything and previously just left the 3D settings as the default NVidia setup, but this caused playback problems after about an hour.

Playback performance is better if I let the 3D application decide and constant reboots are not necessary. I don't think my particular situation is helped by using an NForce 630a motherboard which has very poor NVidia driver support/compatibility.

The main thing is playback is now sorted and I can enjoy superior PQ :)

[madshi]

Does your GPU support CUDA? If so, you can get newest drivers by using the CUDA development drivers which seem to support all GPUs that can do CUDA. I was able to update to 270.32 drivers this way with my NVidia 9400 integrated mainboard.

[bjd]

Yes it is a 2gb GT430 and has 96 CUDA cores. I am currently running the XP 267.24 BETA drivers purely because I was having issues with 266.58 WHQL and thought things might be sorted with newer drivers.

When I get home tonight, I will PM you the madNV12Test results, I have 7600GS on Win7 32bit at work for which I have sent you results.

270.32 drivers on download now thx.

[Thunderbolt8]

Maybe if you force anisotropic filtering or antialiasing that could also affect madVR. It could cost performance, and it could also hurt image quality.would it be possible to create a kind of instruction for ati catalyst and nvidia forceware settings which are recommended to be (de)activated for madvr in terms to reach best quality (and performance)? because I havent much a clue which settings could hurt here.

[mark0077]

would it be possible to create a kind of instruction for ati catalyst and nvidia forceware settings which are recommended to be (de)activated for madvr in terms to reach best quality (and performance)? because I havent much a clue which settings could hurt here.

I think the application can decide what settings to use for anisotropic filter / antialiasing etc but then the nvidia / ati control panel have the ability to override the applications settings which may be the case if a user sets a global preference, (like in my case 16X anisotropic filtering, 4XAA).

All I can think of is if madVR could set the preferred settings, and then after the 3d device / surface is created, check to see if its settings match what madVR asked for (ie if the drivers are forcing anything other than that or not) and perhaps provide some sort of warning, or info in the ctrl-j stats that settings seem to be getting forced by the drivers? At least then the user could know to go and create a custom profile for the app, say mpc-hc, to take settings from the app rather than forcing specific settings itself.. Just a thought...

[madshi]

It's very simple: Set all options to "let the application decide". That should do the trick.

[namaiki]

Weird thing is that I'm pretty sure that is the default setting in the first place.

[mark0077]

Thing is, for people with some of the higher powered gfx cards at least, many like myself are used to setting a global setting like, 16 anisotropic, 4xaa or higher as I said. I'm just thinking alot of those users, when they use madVR might not cop on to the fact that this might negatively effect madVR (I assume it does..)... me included until I read the last few posts :) When madvr creates the 3d device, does it always just get the device in return or is it easy for madVR to know if it got exactly what it asked for, or some variation of it with forced settings like we speak about. I assume it would have to be a manual check on madVR's part if it was to do this check. If its not alot of work, and will improve quality / performance having specific settings, it might be a future improvement for madVR to perform this check and inform the user somehow?

[Thunderbolt8]

It's very simple: Set all options to "let the application decide". That should do the trick.but sometimes I also play older games like wc3 and in game, I cannot set options like aliasing and such. therefore, Id like to have as much enabled as default as possible, with the expections of those settings which should better be disabled for madvr. whichever these should be.

[madshi]

Well, what can I say. I don't know which settings affect madVR and which not. And I don't have the time to find out. But it's easy enough to find out for anyone who wants to know: Just compare the madVR performance (e.g. windowed mode rendering times, dropped frames etc) with various settings set to either "let the application decide" or to "forced enabled". Also you could take screenshots to check whether forcing any options results in a hit in image quality. If you turn off dithering in the madVR settings, screenshots should be bit-for-bit identical. If they're not, then obviously forcing some settings on via GPU control panel harms image quality.

[Razoola]

Razoola. That nvidia cp option never has any effect on my pc. I'm not sure what it does but for me I need that registry tweak to get full range with samsung series 9 - gtx295 - win 7 64 bit

Have you tried it recently. It never used to work for me either but it does now.

[mark0077]

Have you tried it recently. It never used to work for me either but it does now.

Will give it a go this evening, along with some tests on some files that are causing me issues, which I think revolves around timestamps. Will report back on both in a few hours.

[Snowknight26]

madNV12Test seems to exit instantly after pressing OK on this GMA 3150.

[namaiki]

madNV12Test seems to exit instantly after pressing OK on this GMA 3150.

Does madVR even open on the GMA 3150? It doesn't work on the GMA 950 which is based on the same graphics core.

[mark0077]

madshi, one of the mp4 files I'm having issues with frame drops seems to have this issue according to mkvmerge when I was trying to make an mkv from it. No remuxing so far has allowed me to get it to play without evr-cp showing jittery graphs, or madVR dropping frames. ffdshow says its 23.974fps, my display is about 23.962hz~ (nvidia).

"The AVC video track is missing the 'CTTS' atom for frame timecode offsets. However, AVC/h.264 allows frames to have more than the traditional one (for P frames) or two (for B frames) references to other frames. The timecodes for such frames will be out-of-order, and the 'CTTS' atom is needed for getting the timecodes right. As it is missing the timecodes for this track might be wrong. You should watch the resulting file and make sure that it looks like you expected it to."

I assume this is why the frames / timecodes arn't evenly spaced, hence the jittery graphs I get with evr-cp and the fairly frequent frame drops with madVR (maybe 1 every 20 seconds). Do you think this is something you can workaround with the renderer? I can post a link to the file in question if you're interested...

Razoola, I have just clean installed nvidia 267.24 drivers, and of course got the grey / washed out desktop. After changing the option in the control panel you speak of to "Full-screen videos" or "Desktop programs" I get no change at all. I assume this is something being sent to the display.. which my TV at least is ignoring. I have my tv set to display full range, so maybe some tv's have an auto setting also which would "listen" to nvidias output. Not mine anyways. I have to revert to the inf mod