madVR - high quality video renderer (GPU assisted)

[6233638]

That was a good log! The decoder thread is stuck while madVR is trying to read the stream information (audio, video, subtitle tracks) from the splitter. This happens directly after you're restarting playback. This seems to be a splitter bug. Which splitter are you using? Is it the latest version of that splitter? Please list your full filter chain from source to video renderer (only video branch, audio doesn't matter), just to be safe. Thanks.I think the full chain was MPCHC > Nero FLV Splitter > ffdshow > reclock (does it count with video?)

The Nero FLV splitter is from nero 7 ultra which I installed for eac3to, and I used it because some videos wouldnt play with the MPCHC splitter.
With the MPCHC FLV splitter video still freezes after 5s paused but audio works on resume. 0.26 and below is fine with either splitter.


I thought I saw problems with .m2ts from bluray but from testing now it seems to be okay, looks like a FLV problem.

[Gnintendo]

What exactly are the advantages of exclusive mode?

[someone`]

serious judder,delay frame,drop frame when play 1080p movies with ATI Driver other than 10.6 , very smooth with 10.6
Core2Duo 2.0Ghz , ATI 4670 , MPC-HC 1.4.2505 , CoreAVC 2.0 , ffdshow 3562 , Haali Media Splitter 1.10.262.12 , madVR 0.13 --> 0.31
madVR - log.rar - 1.66MB (http://www.zshare.net/download/801736329a1657b1/)

[egandt]

Odd problem with 0.31, which I did not see with 0.30. In ZP 7.1A4 using a 3 minute timeout for the screen saver. Every 3 minutes with 0.31 the screen saver kicks in as if nothing is being played back (AKA Paused). Version 0.30 this does not happen.

ERIC

[6233638]

What exactly are the advantages of exclusive mode?Much better performance, no tearing, no multi-monitor/mixed refresh rate problems. 10-bit output with some cards though not madVR itself yet.

[Snowknight26]

Couple of issues (or.. features) I've run into:

- Going fullscreen in MPC-HC then right clicking to bring up the menu causes madVR to go to windowed for a fraction of a second then back to exclusive mode. Seems as if it's going into fullscreen windowed mode as well.
- Going fullscreen in MPC-HC then switching back to windowed mode caused the chroma plane to shift approximately 60 pixels down. Can't reproduce it though.
- Can't get less than 3 dropped frames when switching modes.
- Had MPC-HC freeze once when trying to open another video. Log (http://stfcc.org/misc/madVR%20-%20log%201.zip).
- Closing the video in MPC-HC doesn't seem to free the file lock on the debug log.

[3x3cUt0r]

Thanks madshi, image quality is amazing.

Still waiting for subtitles support to use it all the time. :thanks:

[namaiki]

Still waiting for subtitles support to use it all the time.

Just use DirectVobSub/VSFilter. I'm quite sure that madshi is very aware of the possibility of having support of MPC-HC's internal subtitle renderer.

[Luv]

Couple of issues (or.. features) I've run into:

- Going fullscreen in MPC-HC then right clicking to bring up the menu causes madVR to go to windowed for a fraction of a second then back to exclusive mode. Seems as if it's going into fullscreen windowed mode as well.
- Can't get less than 3 dropped frames when switching modes.


Same here but with ZP.Don't think these are bugs though.However,even if the "3s switch delay" is checked,the shift wm-fse remains immediate (Almost:3 frame drops)...

[ryrynz]

Madshi, when I switch between exclusive mode and windowed mode and then back to exclusive mode again when MPC loads the next file in the folder automatically after reaching the end of the currently playing file, MPC drops back to windowed mode for a second shows a previous frame then goes back to fullscreen.

Madshi, 0.31 is still doing this for me.

[3ngel]

That error code means "D3DERR_INVALIDCALL". I'm not sure why it occurs. Can you please upload a log for me somewhere (don't attach to this forum, it takes too long for the attachment to be approved).

To create the log, rename "madVR.ax" to something else and then rename "madVR [debug].ax" to "madVR.ax". Then retry to load a video using madVR. After that, zip the "madVR - log.txt" file, which was created on your desktop and upload it somewhere. Thanks.

Hi madshi,

sorry for the late reply.

I tried to create the log in order you to fix, and i've done what you told in order to create it. But the log text doesn't appear on the desktop.

Am i missing something?

Thanks

PS:
Same error with the latest version

[madshi]

I think the full chain was MPCHC > Nero FLV Splitter > ffdshow > reclock (does it count with video?)

The Nero FLV splitter is from nero 7 ultra which I installed for eac3to, and I used it because some videos wouldnt play with the MPCHC splitter.
With the MPCHC FLV splitter video still freezes after 5s paused but audio works on resume. 0.26 and below is fine with either splitter.

I thought I saw problems with .m2ts from bluray but from testing now it seems to be okay, looks like a FLV problem.
Right now I think it's ffdshow. madVR looks for a filter which exposes the interface to ask the audio/video/subtitle streams. If the splitter is not doing that, madVR takes the next filter which does. I guess in your case the FLV splitters don't expose that interface, so madVR probably asks ffdshow, and there have been stability problems before when asking ffdshow about this interface. Can you please try using a different decoder? Just to confirm that ffdshow is causing the problem. Also please try updating to the latest ffdshow build, just to double check. Thanks...

serious judder,delay frame,drop frame when play 1080p movies with ATI Driver other than 10.6 , very smooth with 10.6
Core2Duo 2.0Ghz , ATI 4670 , MPC-HC 1.4.2505 , CoreAVC 2.0 , ffdshow 3562 , Haali Media Splitter 1.10.262.12 , madVR 0.13 --> 0.31
madVR - log.rar - 1.66MB (http://www.zshare.net/download/801736329a1657b1/)
In your 10.5a log, just about everything seems to be very slow. The decoder doesn't seem to be able to keep up, then uploading video frames takes a lot of time. My first suggestion would be to disable the option "upload frames in render thread". According to your log, that should at least bring some improvement. Also you could try to overclock your CPU, just as a test, to see if the decoder is simply too slow. I think newer madVR versions don't handle it very well, if the decoder is too slow. I've no idea why different ATI driver versions should make a difference, though!

Odd problem with 0.31, which I did not see with 0.30. In ZP 7.1A4 using a 3 minute timeout for the screen saver. Every 3 minutes with 0.31 the screen saver kicks in as if nothing is being played back (AKA Paused). Version 0.30 this does not happen.
That's weird! I'm not sure who is responsible for blocking the screen saver. Is it DirectShow? Is it the media player? Or the video renderer? I don't know. madVR has never done anything specific to block the screen saver. So I don't understand why 0.31 would behave any different compared to 0.30. Can you please reboot your PC? Maybe that already fixes the issue?

Couple of issues (or.. features) I've run into:

- Going fullscreen in MPC-HC then right clicking to bring up the menu causes madVR to go to windowed for a fraction of a second then back to exclusive mode. Seems as if it's going into fullscreen windowed mode as well.
This happens for me, too, but only the first time you do that. The 2nd time it should work just fine. Can you confirm that? I might be able to fix that in a future version, but only after I've implemented automatic refresh rate changing etc...

- Going fullscreen in MPC-HC then switching back to windowed mode caused the chroma plane to shift approximately 60 pixels down. Can't reproduce it though.
This has been reported before, but only with the MPC HC internal decoder. Doesn't seem to occur with any other decoder. And personally, I can't reproduce it at all, sadly.

Can't get less than 3 dropped frames when switching modes.
There's nothing I can do about that. Switching takes a bit of time during which frames can't be rendered. Of course I could pause playback during the switching, but I think that would feel even worse, from a user point of view. So I prefer to drop a few frames. I don't think it's a big problem, although of course it would be nice if it wouldn't be necessary.

Had MPC-HC freeze once when trying to open another video. Log (http://stfcc.org/misc/madVR%20-%20log%201.zip).
Thanks! Good log. Seems to be a bug in the madVR finalization code. Can you reproduce that problem? I'm not sure yet if I can fix it with just the log you uploaded. I might need a further log with a future madVR version, but I'm not sure yet.

Closing the video in MPC-HC doesn't seem to free the file lock on the debug log.
Yep, another bug, thanks for reporting!

Madshi, 0.31 is still doing this for me.
I already had commented on that issue here:

http://forum.doom9.org/showthread.php?p=1431170#post1431170

Still waiting for a reply from you.

I tried to create the log in order you to fix, and i've done what you told in order to create it. But the log text doesn't appear on the desktop.

Am i missing something?
Hmmmmm... You did rename "madVR [debug].ax" to "madVR.ax"? And then you restarted the media player? After you did that, do you still see the madVR complaint? Does madVR in this situation show up in the list of filters?

[Mark_A_W]

Mark_A_W,
I think you can configure ZP to not respond to the multimedia keys that MCE remote is sending. I don't use MCE remote but I believe that using a custom media/DVD key file you can remove the "multimedia" keys.
This way you'll avoid the double hits.

Mark,
here is a topic from Zoomplayer's forum that contains a similar request and a solution
http://forum.inmatrix.com/index.php?showtopic=10837



I've no idea about MCE controls and the MCE remote. What would I have to do to make that work?



Thanks Pankov, that's exactly my problem.

And the response on the Inmatrix site, may help madshi :)

From "Ozpa":


In default.key there's this:

// Multimedia Keyboard Basic Functions
//mcAddKey(11,fnNextChapter)
//mcAddKey(12,fnPrevChapter)
//mcAddKey(13,fnStop)
//mcAddKey(14,fnPlay)

(I added the "//"s at start of every line to disable the keys, I could've deleted the whole part all together but I might need to activate them back later on thumbup.gif)

[namaiki]

hahaha. playing a video using madVR in MPC-HC, hiding the caption+menu and resizing the window to be as small as possible makes mpc-hc exit.

[DigitalLF]

MadShi: i get a tearing in "windowed mode" and i have tested to upgrade then downgrade madvr no help there... and then tested upgrading ffdshow and mpc-hc no help there either ...

is there a way to find out where it comes from ?

material is 23.976hz and my screen is a computer screen @ 60hz... and my spec is up to date.

[namaiki]

DigitalLF, how many monitors?

[DigitalLF]

1 monitor... primary... but it has been working before... (does not help to say it but it has...) just a few weeks ago i think...

[Snowknight26]

This happens for me, too, but only the first time you do that. The 2nd time it should work just fine. Can you confirm that? I might be able to fix that in a future version, but only after I've implemented automatic refresh rate changing etc...
Seems to happen all the time but it's not what I'd consider a critical issue. ;)

Thanks! Good log. Seems to be a bug in the madVR finalization code. Can you reproduce that problem? I'm not sure yet if I can fix it with just the log you uploaded. I might need a further log with a future madVR version, but I'm not sure yet.
I could reproduce it maybe 3 times out of 100 or so that I tried, but again, not a really serious issue. I'll be looking forward to further testing the next builds. :)

[madshi]

In default.key there's this:

// Multimedia Keyboard Basic Functions
//mcAddKey(11,fnNextChapter)
//mcAddKey(12,fnPrevChapter)
//mcAddKey(13,fnStop)
//mcAddKey(14,fnPlay)
Not sure what to do with that?

hahaha. playing a video using madVR in MPC-HC, hiding the caption+menu and resizing the window to be as small as possible makes mpc-hc exit.
That's madVR's silent protest! :)

i get a tearing in "windowed mode" and i have tested to upgrade then downgrade madvr no help there... and then tested upgrading ffdshow and mpc-hc no help there either ...

is there a way to find out where it comes from ?

material is 23.976hz and my screen is a computer screen @ 60hz... and my spec is up to date.
1 monitor... primary... but it has been working before... (does not help to say it but it has...) just a few weeks ago i think...
Not sure what has changed. It's possible that you have to modify the flush settings. Can't you use exclusive mode? There should be zero tearing in exclusive mode.

Seems to happen all the time
As in "more often than not"? Or as in "every single time"?

but it's not what I'd consider a critical issue. ;)
It does stop you from being able to use the context menu, does it not? I do consider that a serious bug. Maybe not critical, but serious enough. The big question is if I can change it. madVR itself is not removing the context menu, it's a consequence of Direct3D changing the GPU graphics mode. In my tests I was able to get rid of the problem by doing a manual win32 API ChangeDisplaySettings() call before starting Direct3D rendering. So I hope that this will be the solution to the problem, but on my PC the problem only occurs the first time I try to open the context menu. So maybe the "solution" will not work for you.

Oh wait, there should be a 3 second delay before madVR goes back to exclusive mode, right? So if you are fast enough, you can right click a second time to get the context menu to show up?

I could reproduce it maybe 3 times out of 100 or so that I tried, but again, not a really serious issue. I'll be looking forward to further testing the next builds. :)
Ok, I'll let you know if I need another log with a future version. Maybe not.

[GREG1292]

I have had really no problem with .30 is there any reason
to switch .31? I also have switched to PotPlayer and have
enjoyed the results over MPHC.

Any advantage to using Zoomplayer over PoyPlayer? My picture keeps getting better and better and thought I should
ask instead of troll:thanks:

[noee]

...doing a manual win32 API ChangeDisplaySettings() call

Is this in the .30/.31 release? Is it possible that in dual-display mode, you're setting the refresh on the secondary inadvertently to the primary? For example, my primary is 60Hz and my sec is 24Hz, but for some reason, when entering fse mode, the refresh rate is changing ( I have it disabled in MPC-HC) on my sec monitor.

I assume you're using the EnumDisplaySettings()/ChangeDisplaySettingsEx() API method?

Alright, I'll see if I can get this weird behavior in a log...

[DigitalLF]

Madshi: im sorry... im home sick so today i blame it on that.... yes exclusive work perfectly (sorry) in fullscreen but sometime i watch in "windowed mode" and its only then i get tearing

[Snowknight26]

As in "more often than not"? Or as in "every single time"?
From what I recall (been over 24 hours, cut me some slack :p), it happens every time, but..


It does stop you from being able to use the context menu, does it not? I do consider that a serious bug. Maybe not critical, but serious enough. The big question is if I can change it. madVR itself is not removing the context menu, it's a consequence of Direct3D changing the GPU graphics mode. In my tests I was able to get rid of the problem by doing a manual win32 API ChangeDisplaySettings() call before starting Direct3D rendering. So I hope that this will be the solution to the problem, but on my PC the problem only occurs the first time I try to open the context menu. So maybe the "solution" will not work for you.

Oh wait, there should be a 3 second delay before madVR goes back to exclusive mode, right? So if you are fast enough, you can right click a second time to get the context menu to show up?

At first I thought you and I were talking about something different entirely, but I guess I just explained it poorly.

Right clicking on the video to bring up MPC-HC's menu causes madVR to go from exclusive -> windowed -> exclusive/fullscreen windowed in a very short period of time, but the menu still successfully appears and is functional. As I said before, more of an annoyance than anything. I'll try to get a video of it to clear up the confusion.

[madshi]

I have had really no problem with .30 is there any reason to switch .31?
No. v0.31 does not have any new features, so if v0.30 works perfectly for you, there's no reason to update. You can always look at the changelog to see if there's a new feature that's worth upgrading to.

Is this in the .30/.31 release?
Nope. That was a test I've done on my dev PC, but it's not implemented in madVR yet.

Is it possible that in dual-display mode, you're setting the refresh on the secondary inadvertently to the primary?
No, the latest madVR versions set the refresh rate to 0 when entering fullscreen exclusive mode. That tells Windows to just keep using the same refresh rate. If Windows switches your refresh rate to something else then it's not madVR's fault. All these problems should go away when I implement madVR's automatic refresh rate changing. Other than implementing that there's nothing I can right now. So a log won't help, either.

I assume you're using the EnumDisplaySettings()/ChangeDisplaySettingsEx() API method?
In my dev tests, yes.

[madshi]

At first I thought you and I were talking about something different entirely, but I guess I just explained it poorly.

Right clicking on the video to bring up MPC-HC's menu causes madVR to go from exclusive -> windowed -> exclusive/fullscreen windowed in a very short period of time, but the menu still successfully appears and is functional. As I said before, more of an annoyance than anyhting. I'll try to get a video of it to clear up the confusion.
Ah, I did misunderstand that. Can I have a log of such a multi-back-and-forth switch, please? I've never seen or heard of that effect before!

[noee]

All these problems should go away when I implement madVR's automatic refresh rate changing. Other than implementing that there's nothing I can right now. So a log won't help, either.

Roger. Sounds good, I'll wait for the new feature.

[someone`]

In your 10.5a log, just about everything seems to be very slow. The decoder doesn't seem to be able to keep up, then uploading video frames takes a lot of time. My first suggestion would be to disable the option "upload frames in render thread". According to your log, that should at least bring some improvement. Also you could try to overclock your CPU, just as a test, to see if the decoder is simply too slow. I think newer madVR versions don't handle it very well, if the decoder is too slow. I've no idea why different ATI driver versions should make a difference, though!
much worse when disable "upload frames in render thread" ( both 10.6 and other drivers )
Disable 3dlut help a lot but still drop frame, delay frame every seconds .
With 10.6 , 3dlut or not , performance is the same.
I can't overclock , guess i have to live with 10.6

Video info ( drop frame , delay frame still happen with much lower bitrate 1080p files ) :
Complete name : J:\Clannad\[MII][SS-Eclipse][THORA]_Clannad_OP_(1920x1080)_[DAAF6555].mkv
Format : Matroska
File size : 226 MiB
Duration : 1mn 31s
Overall bit rate : 20.6 Mbps
Movie name : Clannad Opening
Encoded date : UTC 2010-07-06 04:09:01
Writing application : mkvmerge v4.1.1 ('Bouncin' Back') built on Jul 3 2010 22:54:08
Writing library : libebml v1.0.0 + libmatroska v1.0.0
Cover : Yes

Video
ID : 1
Format : AVC
Format/Info : Advanced Video Codec
Format profile : High@L4.1
Format settings, CABAC : Yes
Format settings, ReFrames : 3 frames
Muxing mode : Container profile=Unknown@4.1
Codec ID : V_MPEG4/ISO/AVC
Duration : 1mn 31s
Width : 1 920 pixels
Height : 1 080 pixels
Display aspect ratio : 16:9
Frame rate : 23.976 fps
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Progressive
Title : Clannad OP [THORA]

[DigitalLF]

MadShi: last post on last page.

[Snowknight26]

Ah, I did misunderstand that. Can I have a log of such a multi-back-and-forth switch, please? I've never seen or heard of that effect before!

Here's the log (http://stfcc.org/misc/madVR%20-%20log%202.zip) and just a small correction: happens only once after each fullscreen/windowed mode switch. So it can still be done more than once per video.

Edit: New minor bug (or feature?). When watching, say a 1920x800 video on a 1920x1080 screen in fullscreen mode, hovering the mouse over an area that's not part of the video (so at 1920x1000 for example) causes the madVR seekbar to stay up indefinitely.

[ryrynz]

Still waiting for a reply from you.


Please look into the minimizing just before the new file beings playback. The crash could be unrelated, I haven't experienced enough crashes for me to think it's related to madVR.

I can easily reproduce by playing a file, and maximizing the screen and seeking to the end of the file. Once MPC transitions to the next file in the folder, MPC drops to a normal sized window displays a previously played frame then maximizes again, this only occurs when exclusive mode is enabled.

[oddball]

Is it a lot of work to add support for mpc-hc's internal subtitle display? I don't use ffdshow or VobSub at all because the quality of the subtitles in mpc-hc at desktop res is far superior (Even if mpc's animated subs support is somewhat flakey. Read: dropped frames).

[mark0077]

Hi madshi, I have upgraded my nvidia drivers to the new 260 series... and madvr 0.31. I'm again running into the dropped frame every 26 seconds... Can someone tell me if its expected or not expected to get the frame drop every 26 seconds given the following stats... ie

Display 23.96193hz, movie rate 23.976 fps, clock deviation 0.09999%. If expected then I'm happy enough.. if not then let me know if I can provide any more details. using GTX 295, Win 7 64bit, mpc-hc. The screenshot below taken with print-screen wasn't exactly the print-screen I took, it seems to be the last image in the "buffer?" before I went into exclusive mode...

http://img687.imageshack.us/img687/2534/clockdeviation.th.jpg (http://img687.imageshack.us/i/clockdeviation.jpg/)

[madshi]

much worse when disable "upload frames in render thread" ( both 10.6 and other drivers )
Ok, that's interesting.

Disable 3dlut help a lot but still drop frame, delay frame every seconds .
With 10.6 , 3dlut or not , performance is the same.
I can't overclock , guess i have to live with 10.6

Video info ( drop frame , delay frame still happen with much lower bitrate 1080p files ) :
From the logs I still think that the main problem is a too slow decoder. Do you have the same problems with VC-1 and MPEG2 content?

MadShi: last post on last page.
As I said: It's possible that you have to modify the flush settings.

Please look into the minimizing just before the new file beings playback. The crash could be unrelated, I haven't experienced enough crashes for me to think it's related to madVR.

I can easily reproduce by playing a file, and maximizing the screen and seeking to the end of the file. Once MPC transitions to the next file in the folder, MPC drops to a normal sized window displays a previously played frame then maximizes again, this only occurs when exclusive mode is enabled.
Ok, I'll try to reproduce it.

Is it a lot of work to add support for mpc-hc's internal subtitle display?
As I've been saying on every other page in this thread: subtitle support is on my to do list, but not implemented yet.

Hi madshi, I have upgraded my nvidia drivers to the new 260 series... and madvr 0.31. I'm again running into the dropped frame every 26 seconds... Can someone tell me if its expected or not expected to get the frame drop every 26 seconds given the following stats... ie

Display 23.96193hz, movie rate 23.976 fps, clock deviation 0.09999%. If expected then I'm happy enough.. if not then let me know if I can provide any more details. using GTX 295, Win 7 64bit, mpc-hc. The screenshot below taken with print-screen wasn't exactly the print-screen I took, it seems to be the last image in the "buffer?" before I went into exclusive mode...
It takes quite a while until the clock deviation measurements settle down. Furthermore the measurements start anew every time you pause or stop the video. So the only way to get proper clock deviation measurements is to let the movie run for at least an hour or so - uninterrupted. I think that your clock deviation measurement of ~0.1% is probably incorrect due to a too short measurement time.

If we do the math while ignoring the clock deviation, we get:

24/1.001 - 23.96193 = 0.014094

So your GPU refreshes 0.014094 frames per second less often than it should, if we just look at the system clock. Now if you calculate 1/0.014094 = 70.95. So basically you should have a frame drop every 71 seconds. However, due to the audio clock being different to the video clock (or due to Reclock screwing things up, in case you're using Reclock), this all changes again, so that you may end up with a frame drop every 26 seconds. The solution to the problem should be to create a custom resolution which gets you nearer to the ideal refresh rate of 23.976. But then we still ignore the audio clock deviation. So the ideal refresh rate might be a bit higher or lower than 23.976.

As you can see from the added clock deviation measurements in v0.31, I'm starting to look into the topic of ideal custom resolutions...

[madshi]

Hi madshi, I have upgraded my nvidia drivers to the new 260 series... and madvr 0.31. I'm again running into the dropped frame every 26 seconds... Can someone tell me if its expected or not expected to get the frame drop every 26 seconds given the following stats... ie

Display 23.96193hz, movie rate 23.976 fps, clock deviation 0.09999%. If expected then I'm happy enough.. if not then let me know if I can provide any more details. using GTX 295, Win 7 64bit, mpc-hc. The screenshot below taken with print-screen wasn't exactly the print-screen I took, it seems to be the last image in the "buffer?" before I went into exclusive mode...

http://img687.imageshack.us/img687/2534/clockdeviation.th.jpg (http://img687.imageshack.us/i/clockdeviation.jpg/)
P.S: Just checked the math. If we do this:

24.000 - 23.96193 = 0.03807
1 / 0.03807 = 26.267 frame drops per second

So if the movie would actually be played back in 24.000fps, then you should get exactly 26.267 frame drops per second, which seems to match your experience. So I'm wondering if Reclock modified the playback speed to 24.000fps in your case? That would explain everything. Because if Reclock modifies the video playback speed to 24.000fps, then of course your display must be set to 24.000Hz, too, while it actually is 23.96193Hz. You can trust in that madVR measurement. I'm pretty sure that madVR's display refresh rate measurements are spot on. Not sure on my clock deviation measurements yet, though. That's too new a feature, might still be bugs in that measurement...

[mark0077]

P.S: Just checked the math. If we do this:

24.000 - 23.96193 = 0.03807
1 / 0.03807 = 26.267 frame drops per second

So if the movie would actually be played back in 24.000fps, then you should get exactly 26.267 frame drops per second, which seems to match your experience. So I'm wondering if Reclock modified the playback speed to 24.000fps in your case? That would explain everything. Because if Reclock modifies the video playback speed to 24.000fps, then of course your display must be set to 24.000Hz, too, while it actually is 23.96193Hz. You can trust in that madVR measurement. I'm pretty sure that madVR's display refresh rate measurements are spot on. Not sure on my clock deviation measurements yet, though. That's too new a feature, might still be bugs in that measurement...

Thanks for that. Yeah so it seems my display IS running at 24.000hz when this frame drop issue appears and so its expected.... phew

I see when reclock is running as the audio renderer it detects the display as 24.000 hz and says its adapting media speed to 24.000fps.... yet madVR doesn't display 24.000 anywhere.. Just wondering should madVR also then indicate to me that my display is 24.000 hz or that reclock has changed the rate to 24.000hz. I guess its not standard practice for an audio renderer to change the movie rate.

The sooner I get rid of reclock the better I think :)

[madshi]

Thanks for that. Yeah so it seems my display IS running at 24.000hz when this frame drop issue appears and so its expected.... phew
NO NO NO. You got me totally wrong.

Playback speed seems to be 24.000fps, which means that your decoder delivers 24.000 frames per second to madVR. But your GPU hardware refresh rate seems to be 23.96193Hz, which means that your GPU draws only 23.96193 frames per second. And this mismatch is exactly the cause of the frame drops. madVR is fed with 24.000 frames per second by the decoder, but can only push 23.96193 frames per second to the GPU. Which means that madVR has to drop a frame every 26 seconds to keep audio in sync. Please reread me previous comment. madVR shows the refresh rate correctly. Reclock got it all wrong.

I see when reclock is running as the audio renderer it detects the display as 24.000 hz and says its adapting media speed to 24.000fps....
Yes, and that is exactly the problem. If Reclock were right, we would have a 100% match between movie framerate and display refresh rate. Both would be 24.000. So you would have perfectly fluid playback with no dropped frames. But that's not the case. Reclock changes the playback speed to 24.000, but the real refresh rate is NOT 24.000. That's exactly the reason for the frame drops!

The sooner I get rid of reclock the better I think :)
In this case Reclock increases the frame drops. You would still have frame drops without Reclock, but only one frame every 71 seconds instead of one frame every 26 seconds.

[mark0077]

Sweet! Reclocks display rate detection = bad bad bad. Telling it to keep playback at "original" versus best" = as you say much less drops, thanks for the explanation and apologies if I wasted any more of your time.

Bring on madVR's own refresh rate changing and mpc-hc audio renderers own pal speed down functionality :D

[pankov]

madshi,
I have a few question about the new OSD stats and two feature requests.
I'll start with the requests
1. Can you remove the 3 seconds delay when going from windowed to fullscreen exclusive? If I'm not mistaken it was this way with older versions. I know there is a new option in the settings about these 3 seconds but if I uncheck it I'm not able to show Zoomplayer's popup menu because on destroying the fullscreen exclusive surface the popup menu hides and this makes madVR go back to fullscreen exclusive. For me the best behavior is to have 3 seconds delay between FSW and FSE and no delay when going from Windowed to FSE.

2. (low priority) Since it's not possible to take screenshots in FSE mode using PrnScr can you add some other way of storing/copying the OSD values? May be a shortcut to copy only the values to the clipboard ... or if it's easier to copy the whole frame including the OSD to the clipboard.

And now back to the questions.
I've tried to find the logic of these new stats and I think I did but there is something missing in my math. Here is an example:
1. if I use DirectSound
display rate fluctuates between 23.97539 and 23.97540
source frame rate 23.976
clock deviation 0.00033% (am I right to read this as the (max display rate - min display rate)/max display rate?)
1 frame drop every 29.22 minutes (I guess this is 1 / (movie frame rate - display refresh rate) and depending on the sign it's either drop or delay. Right? In my calculations this should have been between 27.32 and 27.78 minutes)

2. if I use Reclock
display rate fluctuates again between 23.97539 and 23.97540
source frame rate 23.976
clock deviation -0.00246% (why is this so much bigger than before?)
1 frame drop every 13.44 minutes (I'm totally lost how is this calculated ... or what did Reclock change to affect this)

btw
does anybody know how I can configure Reclock to sync the audio to the video without speeding it up to 24.000 from 23.976? ... and I'm not even sure it's doing this because then there will be much more dropped frames ... I guess it's actually changing the rate to 23.975 because it shows "Video hardware -> Refresh rate: 23.975 (DDR)" ... but at the same time shows "Media adaptation -> CINEMA adaptation: media speed changed to 24.000 fps. Audio resampled at 48048Hz"
ooooohhhh - I'm totally lost now.
:(
I simply want everything playing at the exact rate as my monitor (23.97539x) ... or is this possible at all - I mean even madVR to obey the display rate and not drop/delay frames? I believe nobody can feel such small delay/speed up - after all it's 4-5 frames for the whole movie?

[noee]

Pankov:

The msg ""Media adaptation -> CINEMA adaptation: media speed changed to 24.000 fps. Audio resampled at 48048Hz" from Reclock is not entirely accurate. This has been discussed at the Reclock forum and James has just decided to leave it. Basically, if your display rate and frame rate are very close to "exact" (like in your case), then Reclock is not changing the media speed to 24fps.

fwiw, I'm running this config right now, but my numbers are not like yours. I get 3-4 dropped frames per hour, but my deviation is ~0.00482 and it says to repeat 1 frame every ~8.21 minutes. This is for 23.976 source playback @ 23.97692Hz with Reclock resampling. I've had this movie playing in a loop now for hours. And I'm in FS Windowed mode....

HD2600XT, CCC10.3, Win7 x64, Aero OFF, dual-mon (primary 60Hz, sec 23.976Hz, both 1920x1080)
HMS/FFDshow/Reclock/madVR (Windowed-mode)/MPC-HC
3Dlut OFF|Spline64|Spline64|SoftCubic50


Edit (24Hz playback): One other thing I've noticed with Reclock when I have my monitor set to 24Hz (instead of 23.976Hz) and playback 23.976 material. ReClock is definitely altering media speed, and madVR reports the Display Refresh as 24.00089Hz (~ same as Reclock), but madVR shows the movie fps as 23.976 (source filter). I'm guessing that in this case, madVR is reporting the fps from the decoder (FFDshow), even though Reclock is changing the media speed. Methinks this is not ideal if madshi is using these exact numbers for his calcs, but my eyes and ears can't tell if there's a problem, at least for a 2hr movie.

[andybkma]

but if I uncheck it I'm not able to show Zoomplayer's popup menu because on destroying the fullscreen exclusive surface the popup menu hides and this makes madVR go back to fullscreen exclusive. For me the best behavior is to have 3 seconds delay between FSW and FSE and no delay when going from Windowed to FSE.

Hmm, I have no such problems and am also using Zoom Player. Perhaps your answer is not to uncheck that option then? I'm sure some people like that option you are asking to be removed. But of course madshi is the expert...

[dansrfe]

I have perfectly smooth motion using CoreAVC, and upscaling to (1920, x) or (x, 1080) depending on the source A/R in ffdshow raw's AviSynth tab. No Reclock, no exclusive mode, and no 3dLUT. Beautiful rendering speeds, extremely smooth motion, and no tearing whatsoever. Just waiting for internal subtitle support and I will be set with madVR for life :D

[Razoola]

I just wanted to mention the issue causing MPC-HC auto refresh rate change not to work is still the same. This issue only affects exclusive mode however and not windowed mode so I have gone back to using that for the time being.

[ryrynz]

I just wanted to mention the issue causing MPC-HC auto refresh rate change not to work is still the same. This issue only affects exclusive mode however and not windowed mode so I have gone back to using that for the time being.

Madshi is aware of this and plans on implementing auto refresh rate changing through madVR in the near future.

[Neeto]

The microsoft DVD Navigator is usually default DVD Navigator filter.

Does anyone know if using a different DVD Navigator filter e.g. PowerDVD fixes the macromedia error with MadVR?

[DigitalLF]

MadShi: i fixed the tearing by reseting reclock "Clean-up video clock timings database"

[madshi]

1. Can you remove the 3 seconds delay when going from windowed to fullscreen exclusive? If I'm not mistaken it was this way with older versions. I know there is a new option in the settings about these 3 seconds but if I uncheck it I'm not able to show Zoomplayer's popup menu because on destroying the fullscreen exclusive surface the popup menu hides and this makes madVR go back to fullscreen exclusive. For me the best behavior is to have 3 seconds delay between FSW and FSE and no delay when going from Windowed to FSE.
Hmm, I have no such problems and am also using Zoom Player. Perhaps your answer is not to uncheck that option then? I'm sure some people like that option you are asking to be removed. But of course madshi is the expert...
The direct FSE switch when going from true windowed to fullscreen works always for me with MPC HC, but not always with ZP. I think with ZP it works if the movie's aspect ratio matches the screen aspect ratio. ZP seems to go fullscreen in two steps, while MPC HC does it in one. The first ZP switching step is not good enough to allow madVR to go into exclusive mode. That's why there's this 3 second delay. I don't think there's anything I can do about that.

2. (low priority) Since it's not possible to take screenshots in FSE mode using PrnScr can you add some other way of storing/copying the OSD values? May be a shortcut to copy only the values to the clipboard ... or if it's easier to copy the whole frame including the OSD to the clipboard.
Screenshot functionality will come sooner or later.

I've tried to find the logic of these new stats and I think I did but there is something missing in my math. Here is an example:
1. if I use DirectSound
display rate fluctuates between 23.97539 and 23.97540
source frame rate 23.976
clock deviation 0.00033% (am I right to read this as the (max display rate - min display rate)/max display rate?)
Clock deviation is not what you seem to think it is. Clock deviation is a measurement of how much the *reference clock* deviates from the *system clock*. The reference clock is usually provided by the audio renderer. The MS audio renderers take the clock from the audio hardware.

Basically, if you have a movie framerate of 23.976fps and a display refresh rate of 23.976Hz, you would think that there should not be any frame drops/repeats. But that doesn't actually have to be true. If the audio hardware's clock deviates from the system clock, this must be taken into consideration, too.

A "clock deviation" measurement of 0.00033% means that the reference clock runs 0.00033% faster than the system clock.

I get 3-4 dropped frames per hour, but my deviation is ~0.00482 and it says to repeat 1 frame every ~8.21 minutes. This is for 23.976 source playback @ 23.97692Hz with Reclock resampling.
There's a bug in my math. I think the clock deviation is measured correctly, but the estimate of how many frames are dropped/repeated is wrong. Let's do the math manually:

display refresh rate: 23.97692 Hz
movie frame rate: 23.976 fps
clock deviation: 0.00482%

expected frame drops/repeats per second: 1 / ((23.976 + 0.00482%) - 23.97692) = 4244

Since the corrected movie frame rate (23.976 + 0.00482%) is higher than your display refresh rate, we should expect frame drops and not frame repeats. So you should have about 1 frame drop every 4244 seconds. Which is about 1 frame drop every 71 minutes. Not sure why you have 3-4 frame drops per hour. Maybe the clock deviation measurement is not exact enough yet.

Edit (24Hz playback): One other thing I've noticed with Reclock when I have my monitor set to 24Hz (instead of 23.976Hz) and playback 23.976 material. ReClock is definitely altering media speed, and madVR reports the Display Refresh as 24.00089Hz (~ same as Reclock), but madVR shows the movie fps as 23.976 (source filter). I'm guessing that in this case, madVR is reporting the fps from the decoder (FFDshow), even though Reclock is changing the media speed. Methinks this is not ideal if madshi is using these exact numbers for his calcs, but my eyes and ears can't tell if there's a problem, at least for a 2hr movie.
The framerate reported by the decoder needs to be adjusted by the clock deviation. If you let Reclock change playback speed from 23.976 to 24.000Hz, the audio clock should run faster than the system clock by the factor of 24/23.976, so the clock deviation measurement should be roughly 0.1%.

MadShi: i fixed the tearing by reseting reclock "Clean-up video clock timings database"
Good to hear.

[mark0077]

madshi, not sure if you're aware of this, but the issue of the right click menu not appearing in exclusive still remains (for me in wind7 64bit anyways). Is this fixable from madvr's end of things.

I was able to workaround by rightclicking a second time, but now that I have set madVR to enter exclusive mode immediately (instead of waiting for 3 seconds), multiple right clicks only makes madVR stutter for a second, then immediately reenter exclusive fullscreen mode again... so I'm unable to get a rightclick menu in this configuration to change audio streams / do other stuff. Unsure if you can do anything about this, not a big issue anyways.

[leeperry]

:thanks: for 0.31 madshi, my dropped frames problem a few secs after seeking seems to have been completely solved! and for fun I was comparing your VR to EVR, hah :D

even when using Leak's chroma upsampler, mVR looks noticeably sharper and clearer to my eyes..maybe it's PotPlayer's that's using a funky HLSL scaler from Haali, I know KMP was(even in 1:1)..so ya never know, but a super soft chroma allows to get a sharper luma anyway as they don't interfere w/ each other anymore.
madVR is reporting the fps from the decoder (FFDshow), even though Reclock is changing the media speed. Methinks this is not ideal if madshi is using these exact numbers for his calcs, but my eyes and ears can't tell if there's a problem, at least for a 2hr movie.
Reclock works behind mVR's back, the latter cannot know when a 25 or 23.976fps stream is being reclocked to 24fps...all those stats are irrelevant when using Reclock.

I've seen fancy stats like I should be getting a duped frame every 20s, but I've never seen any...nor did I expect to :p
i fixed the tearing by reseting reclock "Clean-up video clock timings database"
It's good practice to reset Reclock's timings when updating mVR, your media player or your graphic drivers...I've had the same kind of hiccuping problems before doing it. It makes sense as madshi is always finetuning his VR timings, and Reclock needs to adapt accordingly.

[madshi]

madshi, not sure if you're aware of this, but the issue of the right click menu not appearing in exclusive still remains (for me in wind7 64bit anyways). Is this fixable from madvr's end of things.
I hope it will go away when I have implemented auto refresh rate changing.

Reclock works behind mVR's back, the latter cannot know when a 25 or 23.976fps stream is being reclocked to 24fps...all those stats are irrelevant when using Reclock.
Actually madVR v0.31 does see what Reclock is doing. That is the new OSD information "clock deviation". E.g. if you do 25 -> 23.976, the clock deviation info should show ca. -4%. The estimate of how many frames will be dropped/repeated per second is incorrect in v0.31, though.

[buzzqw]

I hope it will go away when I have implemented auto refresh rate changing.

:eek: any ETA ?

BHH