madVR - high quality video renderer (GPU assisted)

[wanezhiling]

Oh. I can reproduce it. Weird, will have to double check...

:) Thanks for confirming it.

[Dodgexander]

Just a couple of things, what kind of upscaling should a Radeon 6870 handle? Seems Jinc4 is the best you can get.

Another thing, when using exclusive mode, with my secondry monitor sometimes applications get "stuck". They fade, and become unresponsive until I take madvr out of exclusive mode by right clicking on mpc. Is there any reason this is happening? Seems a driver thing.

Disabling desktop composition does the trick, but is there away to keep it without this happening?

Disabling exclusive mode results in dropped frames sadly...

[Jong]

Hi Madshi, just looking at your new "repeated frames" counter.

When playing R1 DVDs @59.94Hz with JRiver MC's Videoclock I get 12 repeated frames counted every second. See attached screenshot.

http://sdrv.ms/Tr1d3C

The clock deviation eventually converges on 0.00000%. In the link the rendering time is pretty close to the frame time, but I have repeated with more basic upscaling, so rendering takes <5ms and still the same 12 repeated frames a second. Seems it is not caused by over-processing.

Can't say I can notice them so maybe it is an error in the OSD? On the other hand, with so many, they may be hard to spot combined with pulldown judder. On the other, other hand maybe it is just counting the pulldown frames?! If the latter, I guess it could also be intentional, although a little confusing if it is.

Of course I can do a log, but it might be better for you to see if you see the same with a R1 DVD.

[madshi]

ok madshi how about this?
link (http://www.tigerdirect.com/applications/SearchTools/item-details.asp?EdpNo=4501123&CatId=7387)
Looks much better.

Just a couple of things, what kind of upscaling should a Radeon 6870 handle? Seems Jinc4 is the best you can get.
Jinc3 AR is all you need. No need to use higher taps for Jinc.

Another thing, when using exclusive mode, with my secondry monitor sometimes applications get "stuck". They fade, and become unresponsive until I take madvr out of exclusive mode by right clicking on mpc. Is there any reason this is happening? Seems a driver thing.

Disabling desktop composition does the trick, but is there away to keep it without this happening?
Not sure why that is happening. I don't remember anyone else reporting this, so it might be something specific to your PC/installation somehow.

Hi Madshi, just looking at your new "repeated frames" counter.

When playing R1 DVDs @59.94Hz with JRiver MC's Videoclock I get 12 repeated frames counted every second. See attached screenshot.

http://sdrv.ms/Tr1d3C

The clock deviation eventually converges on 0.00000%. In the link the rendering time is pretty close to the frame time, but I have repeated with more basic upscaling, so rendering takes <5ms and still the same 12 repeated frames a second. Seems it is not caused by over-processing.

Can't say I can notice them so maybe it is an error in the OSD? On the other hand, with so many, they may be hard to spot combined with pulldown judder. On the other, other hand maybe it is just counting the pulldown frames?!

Of course I can do a log, but it might be better for you to see if you see the same with a R1 DVD.
Ah, I already know where this is coming from. If you watch a soft-telecined DVD with DXVA deinterlacing, the video stream really only has 24 encoded frames per second. madVR does double rate deinterlacing, so this is practically doubled to 48 output frames. 60-48 = 12 frames. So basically because the DVD is soft-telecined, there are only 48 deinterlaced frames per second. But the refresh rate is 60Hz, so there are 12 vsync intervals which have no unique video frames, so you get 12 repeated frames per second. Makes perfect sense. I guess in this situation the repeated frame counter isn't really useful. I'm not sure if I can easily detect this. Of course the ideal solution would be to use madVR's forced film mode.

[huhn]

i'm sure u can fix this issue but i still want to point it out:

running a 1440x1080p60i h264 ar 16/9 with dxva cp and lavfilter yadif (50/60p) my grafikcard 6770 can't handle it even with bilinear filtering. ofc the card is more then fast enough but when using dxva the card stays at 400 mhz and not the 850 mhz max power state. even when i pause the play back and start a computer game the grafikcard will still stay at 400 mhz.

without dxva cp i can use jinc 3 ar for both chroma and luma and i'm still under 14 ms render time so the card is just fast enough.

when low power grafikcards like the 6450 behaver like this there is noway for a system with the need of dxva to do it with madvr. but without dxva is should be possible.

[madshi]

@huhn, AMD cards are known to be slow with copyback (especially everything older than 7xxx). So simply don't use DXVA Copyback. Use either Native DXVA or don't use DXVA at all. There's nothing I can do about it. It's not a problem in madVR. Actually the slowness is caused inside of LAV Filters when doing the Copyback.

[Jong]

Madshi, further on repeated frames, just been playing a PAL DVD @50Hz (50.0011Hz) with Videoclock. I was surprised to see 2 repeated frames in about 20 mins. Stats say 1 frame repeat every 15 mins even with 0.00001% clock deviation. Any thoughts why this would be?

For blu-ray and mkv 720p, even sped up from 24p to 25p, the repeat rate in stats is in days (or occasional many hours).

[huhn]

are u sure the slowness is a cause of lav? it's the "same" with native(i have to encode the video in 60p to test it but the mhz still stays at 400) the card just stays at 400 mhz is not like amd is slow amd is simpling blocking the potential of the card when dxva is in used. at least that is what i see there.

but yeah there is nothing u can do about so i leave it at this maybe amd "fix" this...

[madshi]

Madshi, further on repeated frames, just been playing a PAL DVD @50Hz (50.0011Hz) with Videoclock. I was surprised to see 2 repeated frames in about 20 mins. Stats say 1 frame repeat every 15 mins even with 0.00001% clock deviation. Any thoughts why this would be?
The math is pretty simple: 50.0011Hz is 0.0011Hz too high. If you do the math, this results in 0.066 frame repeats per minute. Or one frame repeat every 15 minutes. If the clock deviation is pretty much zero, that's what it is. Try without Videoclock. It might actually result in less frame repeats. If the clock deviation is > 0% then it might counter the too-high refresh rate.

are u sure the slowness is a cause of lav? it's the "same" with native(i have to encode the video in 60p to test it but the mhz still stays at 400) the card just stays at 400 mhz is not like amd is slow amd is simpling blocking the potential of the card when dxva is in used. at least that is what i see there.

but yeah there is nothing u can do about so i leave it at this maybe amd "fix" this...
I don't know about why the clock stays at 400. But apart from that it's a well known problem that AMD GPUs are notoriously slow with Copyback. It's less of a problem with 7xxx generation GPUs and with integrated CPU/GPU, but with older dedicated GPUs it definitely is a problem.

[DragonQ]

I'm sure DXVA2 Native works fine on AMD. IIRC the reason for DXVA2 Copy-Back not working well enough on older AMD cards is simply to do with their architecture.

[noee]

@huhn: You could try setting up a profile in CCC, then modifying the profile to "disable" the power mgt. Numerous examples out there...

[Jong]

The math is pretty simple: 50.0011Hz is 0.0011Hz too high. If you do the math, this results in 0.066 frame repeats per minute. Or one frame repeat every 15 minutes. If the clock deviation is pretty much zero, that's what it is. Try without Videoclock. It might actually result in less frame repeats. If the clock deviation is > 0% then it might counter the too-high .I guess I don't understand why this happens with DVD but not with .mkv or even .vob, played back at the same refresh. Yes, I know DVD and file playback are different (nevcairiel, if you are reading ;) ) but does that have to mean one will be worse? It looks like Videoclock is not doing its job here and I thought MadVR and Videoclock were supposed to cooperate to eliminate judder from small differences like this :confused:

[DragonQ]

@huhn: You could try setting up a profile in CCC, then modifying the profile to "disable" the power mgt. Numerous examples out there...

It's also a good idea to make sure the various related items in Windows' Power Options are set properly - it was impossible for me to get smooth playback in MadVR on my new laptop until I did this. The two big ones are Multimedia Settings -> When Playing Video and Intel Graphics Settings (when applicable).

[Prinz]

Not sure why that is happening. I don't remember anyone else reporting this, so it might be something specific to your PC/installation somehow.

I have the same Problem. It been always this way with exclusive mode on my PC. (ATI 2600 XT - Win 7 x64)

[madshi]

I guess I don't understand why this happens with DVD but not with .mkv or even .vob, played back at the same refresh.
Well, what does the OSD say when doing mkv/vob playback?

It looks like Videoclock is not doing its job here and I thought MadVR and Videoclock were supposed to cooperate to eliminate judder from small differences like this :confused:
madVR and VideoClock don't "work together". VideoClock asks madVR about which refresh rate it has measured. That's the only thing where they talk to each other. I don't know the exact design of VideoClock. You'd have to ask the JRiver guys about that.

[Dodgexander]

I have the same Problem. It been always this way with exclusive mode on my PC. (ATI 2600 XT - Win 7 x64)

Looks much better.

Not sure why that is happening. I don't remember anyone else reporting this, so it might be something specific to your PC/installation somehow.


Ill try get a screenshot next time it happens. Perhaps its related to other applications using the dxva, flash for example?

[Prinz]

On my System the stuck window is always clearly stuck in Fading away (window is semi transparent).

And I can most of the time unstuck the window, if I do the following:

1. Right click on taskbar and start the taskmanager.
2. Clicking at least twice on the Taskmanager Icon (so that is minimized and restored at leat once)
3. with Taskmanager restored, close it.

Most of the time then also the stuck window disappears too (sometimes after 2-3 seconds). Sometimes I have to do this more then once to unstuck the windows.

[Jong]

Well, what does the OSD say when doing mkv/vob playback?


madVR and VideoClock don't "work together". VideoClock asks madVR about which refresh rate it has measured. That's the only thing where they talk to each other. I don't know the exact design of VideoClock. You'd have to ask the JRiver guys about that.OK Madshi, thanks. I think in getting it clearer in my head now.

So you would expect some clock deviation if Videoclock is correcting for a refresh rate error right? When I play a .vob I get a clock deviation of .0022% and a repeat rate in days. When I play the DVD the clock deviation converges on zero and I get a much higher error rate. Seems MC is not enabling Videoclock for DVD.

[Dodgexander]

On my System the stuck window is always clearly stuck in Fading away (window is semi transparent).

And I can most of the time unstuck the window, if I do the following:

1. Right click on taskbar and start the taskmanager.
2. Clicking at least twice on the Taskmanager Icon (so that is minimized and restored at leat once)
3. with Taskmanager restored, close it.

Most of the time then also the stuck window disappears too (sometimes after 2-3 seconds). Sometimes I have to do this more then once to unstuck the windows.

For me just right clicking on the video does the trick, since it takes it out of exclusive mode, then everything flushes and goes back to normal again, until it next happens..

[madshi]

So you would expect some clock deviation if Videoclock is correcting for a refresh rate error right? When I play a .vob I get a clock deviation of .0022% and a repeat rate in days. When I play the DVD the clock deviation converges on zero and I get a much higher error rate. Seems MC is not enabling Videoclock for DVD.
Yes, seems like that.

[YxP]

Disabling exclusive mode results in dropped frames sadly...

I had problems with windowed mode as well. Minimizing other windows (especially if you have facebook open) on your first screen might help, at least it worked for me. Using windowed mode now myself :)

[aufkrawall]

Weird thing:
After some time I'm getting framedrops when I play this video:
http://www.oc-burner.de/ftp/Videos/juddertest/juddertest_60.mp4

Display is at 60.02 Hz, so it is higher than video fps.
Any idea why framedrops do occur?
Already played around with buffer sizes etc.

[Prinz]

For me just right clicking on the video does the trick, since it takes it out of exclusive mode, then everything flushes and goes back to normal again, until it next happens..

I mean it's a way without leaving the exclusive mode. Of course leaving the exclusive mode always works too.

[huhn]

using cpu everything is fine and the full power of the gpu is used
max power state all fine
cpu decoing = http://s3.imgimg.de/uploads/cpu130a761fpng.png

using cp or native results in frame drop. because the gpu runs at 400 mhz max

native = http://s3.imgimg.de/uploads/nativeb0ed5fedpng.png
copy back = http://s3.imgimg.de/uploads/copyback4e8ec518png.png

i'm posting this to show that using dxva ist stopping the gpu from entering max power state with is bad for madvr that is all

by the way cp works a lot better in this case and that doesn't make sense at all.

I'm sure DXVA2 Native works fine on AMD. IIRC the reason for DXVA2 Copy-Back not working well enough on older AMD cards is simply to do with their architecture.

like u can see it didn't and cp is not much slower for me

@huhn: You could try setting up a profile in CCC, then modifying the profile to "disable" the power mgt. Numerous examples out there...

doing this mpc hc crashed or the cpu clock still goes down to 400 mhz. i only managed to get it running with cp and only sometimes.

i simple use my cpu and every thing is fine is not like this can be changed anyway

tl;dr
dxva max 400 mhz cpu max 850 that's all

[Jong]

Playing a R1 DVD (same soft telecined DVD as earlier) @50Hz using MC I get a reported clock deviation of 0%, but very few drop/repeated frames (none in a few minutes). The "x repeat frames every.." line is completely missing from the OSD. Does this make sense? If I play a 24p mkv I get >4% clock deviation, which is more what I would expect. But with no clock deviation and such a big frame rate/refresh rate difference shouldn't we be seeing frames repeated like crazy? And where has the "x repeat frames every" line gone?

[Dodgexander]

http://imageshack.us/scaled/thumb/834/example2fh.png (http://imageshack.us/photo/my-images/834/example2fh.png/)
http://imageshack.us/scaled/thumb/571/exampleqi.png (http://imageshack.us/photo/my-images/571/exampleqi.png/)

Examples of the ghosting images you get on a second screen in exclusive mode.

[Niyawa]

Personally i prefer copy-back, because its just less hassle and allows all madVR features, like IVTC which currently doesn't work with Native.
On some GPUs madVR will even do its own "copy-back" internally because its currently the only way for it to get maximum quality, but i think thats mostly limited to NVIDIA, IIRC. In this case, copy-back or CUVID is even more to be recommended, as the performance penality is nearly the same, and using cb/cuvid gives you the flexibility again.

However, you should only use Copy-Back if you have a SSE 4.1 capable CPU (introduced with Penryn Core2 models, on the AMD side only with Bulldozer)
I have a Core 2 Duo so I'm good to go, thanks for the wonderful answer.

Thats to be expected, and part of the whole idea that is copy-back. With copy-back, all DXVA stuff stays internal to the decoder, and the outside world has no clue that DXVA is being used - which means it has the full flexibility of a software decoder and none of the DXVA Native limitations.
Now it makes more sense. Thanks again.

Which bug do you mean? I don't remember...
It's not really a bug. You mentioned it here (http://forum.doom9.org/showpost.php?p=1606007&postcount=16326) and here (http://forum.doom9.org/showpost.php?p=1606030&postcount=16331).

[yahyoh]

thanks a lot madshi for the new release :thanks:

now frame skipping will not make mpc stuck for 2-3 sec :)

but when using frame skipping (many frames together) the image don't move until i remove my finger from the skipping key :(

[madshi]

Weird thing:
After some time I'm getting framedrops when I play this video:
http://www.oc-burner.de/ftp/Videos/juddertest/juddertest_60.mp4

Display is at 60.02 Hz, so it is higher than video fps.
Any idea why framedrops do occur?
Already played around with buffer sizes etc.
What does the madVR OSD say in the moment when the frame drops occur? If possible, make a screenshot of the OSD in that moment.

Playing a R1 DVD (same soft telecined DVD as earlier) @50Hz using MC I get a reported clock deviation of 0%, but very few drop/repeated frames (none in a few minutes). The "x repeat frames every.." line is completely missing from the OSD. Does this make sense? If I play a 24p mkv I get >4% clock deviation, which is more what I would expect. But with no clock deviation and such a big frame rate/refresh rate difference shouldn't we be seeing frames repeated like crazy? And where has the "x repeat frames every" line gone?
IIRC, the "x repeat frames every" line only appears if movie framerate and display refresh rate are reasonably close. The DVD is 60i in its original form and the refresh rate is anything even remotely close to 60i, so that's why the "x repeat frames every" line is not showing. I don't understand the clock deviation behaviour, though. When playing 24p, it makes sense that it's around 4% because 24p -> 25p is a bit more than 4% difference. But in the first case you said clock deviation was at 0%. That doesn't make sense if the source is 60i and the refresh rate is 50Hz. Also in that case it doesn't make sense that you saw repeated frames in the OSD. You should not have.

Examples of the ghosting images you get on a second screen in exclusive mode.
I don't think this is madVR's fault. madVR might make this problem appear, but I think it's actually a bug in the OS or in the GPU drivers. Maybe I can work around it, but currently I can't reproduce it. Maybe some time in the future.

It's not really a bug. You mentioned it here (http://forum.doom9.org/showpost.php?p=1606007&postcount=16326) and here (http://forum.doom9.org/showpost.php?p=1606030&postcount=16331).
v0.85.8 should at least make madVR behave similar to other renderers, when using IVTC with a 60Hz display mode. The "final" solution for 24fps @ 60Hz is still some time away. But other renderers don't have a real solution for this right now, either.

[madshi]

but when using frame skipping (many frames together) the image don't move until i remove my finger from the skipping key :(
You mean you keep the frame skipping key pressed? This works for me, the video visually plays when I do that on my PC with MPC-HC.

[yahyoh]

You mean you keep the frame skipping key pressed? This works for me, the video visually plays when I do that on my PC with MPC-HC.

yep

i'm using the right arrow key for frame skipping so if try to skip many frames at the same time the video frames will stuck at the same frame until i remove my finger from the key :(

i'm using latest madvr + latest mpc + latest lav filter :rolleyes:

[DragonQ]

What would be the final solution for 24p @ 60 Hz then?

1, 1, 1-2 (blended), 2, 2, 3, 3, 3-4 (blended), 4, 4.... ?

[pie1394]

What would be the final solution for 24p @ 60 Hz then?

1, 1, 1-2 (blended), 2, 2, 3, 3, 3-4 (blended), 4, 4.... ?

I am curious if this would create ghost image effect during the transition while motion blur effect has already existed in the 24-fps FILM-captured images...

[Niyawa]

v0.85.8 should at least make madVR behave similar to other renderers, when using IVTC with a 60Hz display mode. The "final" solution for 24fps @ 60Hz is still some time away. But other renderers don't have a real solution for this right now, either.
I understand. Could you explain in more detail exactly what makes this complicated to deal with? I want to make sure I have a solid answer if people ask me. Oh yeah just one more thing. When I open the madVR OSD with ReClock, my "clock deviation" is always between 0.30218 or similar. The moment I start using DirectSound instead it goes back to 0.00001. Do you know what could cause this to happen? It never gave me any problems, but I'm curious about it. Did some search about what "clock deviation" is and never found an answer too.

[Oline 61]

On madVR 0.85.7 I was using display modes "1920x1200p24, 1920x1200p25, 1920x1200p30" and everything was working fine.
Version 0.85.8 no longer switches to 1920x1200p30 for 30 fps material and instead uses 1920x1200p24. Switching back to 0.85.7 fixes this.

System is Windows 7 on old Athlon 64 and nVidia 9400GT with a Samsung T240HD monitor connected via DVI.

[Jong]

IIRC, the "x repeat frames every" line only appears if movie framerate and display refresh rate are reasonably close. The DVD is 60i in its original form and the refresh rate is anything even remotely close to 60i, so that's why the "x repeat frames every" line is not showing. I don't understand the clock deviation behaviour, though. When playing 24p, it makes sense that it's around 4% because 24p -> 25p is a bit more than 4% difference. But in the first case you said clock deviation was at 0%. That doesn't make sense if the source is 60i and the refresh rate is 50Hz. Also in that case it doesn't make sense that you saw repeated frames in the OSD. You should not have.My mistake. I was looking for one of the "dropped/repeated/delayed frame" counters to increase but they didn't. the reason, of course, is because the repeated frame counter is not even there for the same reason the "x repeat frames every" is missing - because MadVR thinks the frame rates are too far apart.

Of course, because of soft-telecine, in fact they are not too far apart and it would be useful (particularly for me with my current MC issue) to see the repeat counter in this case, but I get that is a tricky issue.

At least this means that R1 DVDs are most likely consistent with R2 DVDs in MC, i.e. Videoclock is not currently working, since both show a deviation of 0% for DVD, but a proper corrected figure for .vob. Just a shame we can't see the repeat counter clocking up fast in the R1 case.

[madshi]

i'm using the right arrow key for frame skipping so if try to skip many frames at the same time the video frames will stuck at the same frame until i remove my finger from the key :(

i'm using latest madvr + latest mpc + latest lav filter :rolleyes:
I'm not sure, but this might have to do with the speed of your PC. Are you using DXVA or CUVID decoding? Try software decoding. Or try using Bilinear for everything, just to check if that makes the problem go away.

What would be the final solution for 24p @ 60 Hz then?

1, 1, 1-2 (blended), 2, 2, 3, 3, 3-4 (blended), 4, 4.... ?
Yes.

I am curious if this would create ghost image effect during the transition while motion blur effect has already existed in the 24-fps FILM-captured images...
I don't see why there would be a ghost image effect. Frame blending is used all the time to bring old NTSC 60i TV series to PAL 50i TV and it's not "visible" during motion. Of course when using it to reduce the fps, it reduces quality by making everything quite a bit softer. But that problem shouldn't apply when increasing fps significantly. I expect a proper 24 -> 60 solution to look almost identical to true 24 -> 24 playback.

I understand. Could you explain in more detail exactly what makes this complicated to deal with? I want to make sure I have a solid answer if people ask me. Oh yeah just one more thing. When I open the madVR OSD with ReClock, my "clock deviation" is always between 0.30218 or similar. The moment I start using DirectSound instead it goes back to 0.00001. Do you know what could cause this to happen? It never gave me any problems, but I'm curious about it. Did some search about what "clock deviation" is and never found an answer too.
There are 3 clocks during video playback:

(1) System clock ("time").
(2) Video refresh rate.
(3) Audio hardware clock.

The system clock really doesn't have anything to do with video playback, but madVR uses it to measure and compare the other two clocks. Now it is possible (even probable) that the system clock is not perfect. But that's not a big problem because if it's e.g. slightly too fast, both video and audio clocks will be measured wrong in the exact same way, so everything's fine.

If the video and audio hardware clocks are "perfect", there should be no dropped or repeated frames (if your PC is fast enough). However, video and audio clocks are usually both imperfect. If both are imperfect in the same way (e.g. being 0.1% too fast), again there's no problem. But if the amount of "deviation from perfectness" of audio and video clocks is not identical then madVR has to either drop frames (video clock slower than audio clock) or repeat frames (video clock faster than audio clock) to make sure video and audio stay in sync.

The refresh rate as shown in the first line in the madVR OSD is madVR's measurement of the video clock, measured by using the system clock. The audio "deviation" in the madVR OSD shows how much the audio clock deviates from "perfectness", measured by the system clock. If e.g. the measured refresh rate is 24.24000Hz (1% too fast for 24.000 content) and the audio deviation is exactly 1%, too, then there should be no frame drops/repeats. If the video/audio deviation differs, there have to be drops or repeats.

Reclock replaces the audio clock. Through real-time audio resampling Reclock makes sure that the audio clock exactly matches the video clock. This way madVR doesn't have to drop or repeat frames. E.g. if the video clock is 0.1% too fast then Reclock makes sure that the audio clock is also 0.1% too fast. Things get more complicated if you ask Reclock to change the playback fps. In that case I would expect the measured audio clock deviation to be much bigger. E.g. if you play 24p content at 25/50Hz I would expect the audio deviation measurement to show a roughly 4% deviation because Reclock achieves the fps change by running the audio clock much faster or slower.

On madVR 0.85.7 I was using display modes "1920x1200p24, 1920x1200p25, 1920x1200p30" and everything was working fine.
Version 0.85.8 no longer switches to 1920x1200p30 for 30 fps material and instead uses 1920x1200p24. Switching back to 0.85.7 fixes this.

System is Windows 7 on old Athlon 64 and nVidia 9400GT with a Samsung T240HD monitor connected via DVI.
Hmmmm... I've changed the logic a bit in v0.85.8. In earlier versions 60i content was treated as needing 30Hz, unless DXVA deinterlacing was turned on. In that case it was treated as needing 60Hz. In v0.85.8 60i content is always treated as needing 60Hz, unless madVR knows for sure that DXVA deinterlacing won't be needed. So my best guess is that madVR asks for a 60p mode, but doesn't find it in your list. I'm not sure why it then ends up with 24p. I guess it's because 60i content can sometimes be IVTCed back to 24p. But I guess it would make more sense to use 30p in this situation. I'll put that on my to do list for the next build.

However, I'm wondering if your list of resolutions really makes sense. There is only very very few real 30p content. Most is either film content (after IVTC 24p) or video content (after DXVA deinterlacing 60p). Is your display not able to accept 1080p60?

[Niyawa]

There are 3 clocks during video playback:

(1) System clock ("time").
(2) Video refresh rate.
(3) Audio hardware clock.

The system clock really doesn't have anything to do with video playback, but madVR uses it to measure and compare the other two clocks. Now it is possible (even probable) that the system clock is not perfect. But that's not a big problem because if it's e.g. slightly too fast, both video and audio clocks will be measured wrong in the exact same way, so everything's fine.

If the video and audio hardware clocks are "perfect", there should be no dropped or repeated frames (if your PC is fast enough). However, video and audio clocks are usually both imperfect. If both are imperfect in the same way (e.g. being 0.1% too fast), again there's no problem. But if the amount of "deviation from perfectness" of audio and video clocks is not identical then madVR has to either drop frames (video clock slower than audio clock) or repeat frames (video clock faster than audio clock) to make sure video and audio stay in sync.

The refresh rate as shown in the first line in the madVR OSD is madVR's measurement of the video clock, measured by using the system clock. The audio "deviation" in the madVR OSD shows how much the audio clock deviates from "perfectness", measured by the system clock. If e.g. the measured refresh rate is 24.24000Hz (1% too fast for 24.000 content) and the audio deviation is exactly 1%, too, then there should be no frame drops/repeats. If the video/audio deviation differs, there have to be drops or repeats.

Reclock replaces the audio clock. Through real-time audio resampling Reclock makes sure that the audio clock exactly matches the video clock. This way madVR doesn't have to drop or repeat frames. E.g. if the video clock is 0.1% too fast then Reclock makes sure that the audio clock is also 0.1% too fast. Things get more complicated if you ask Reclock to change the playback fps. In that case I would expect the measured audio clock deviation to be much bigger. E.g. if you play 24p content at 25/50Hz I would expect the audio deviation measurement to show a roughly 4% deviation because Reclock achieves the fps change by running the audio clock much faster or slower.
Thanks for the wonderful answer. I better bookmark this so I won't forget where to look up when I want more details. Let's just see if I got this. Basically the problem with 24p@60Hz is that you need to find a way to make the Video and Audio clock to stay sync without the need of repeated or dropped frames. That's hard to do with ReClock because of the way it works (by changing not only the refresh rate, but the video fps as well). I may have got it wrong since this is all news to me but for now that's enough, thanks again.

[madshi]

Let's just see if I got this. Basically the problem with 24p@60Hz is that you need to find a way to make the Video and Audio clock to stay sync without the need of repeated or dropped frames. That's hard to do with ReClock because of the way it works (by changing not only the refresh rate, but the video fps as well). I may have got it wrong since this is all news to me but for now that's enough, thanks again.
Wait. My answer was a reply to your question about what that "deviation" information in the OSD means. 24p@60Hz is a different topic. The simple problem with 24p@60Hz is that 60 is not a simple mulitply of 24, which means that every renderer out there today shows a 3:2 motion judder. Basically frame A is shown for 2 vsync periods, then frame A+1 is shown for 3 vsync periods, then frame A+2 for 2 vsyncs, frame A+3 for 3 vsyncs, etc. This uneven frame repeat pattern results in that motion isn't perfectly smooth.

[pankov]

Not sure why that is happening. I don't remember anyone else reporting this, so it might be something specific to your PC/installation somehow.
I think I've reported this a long time ago but couldn't find it now. It happens almost every time I try to open a window on my primary display while using exclusive mode on the secondary. It happens both when staring a new application or just restoring a previously minimized one. It also happens when opening new windows in already running app. If you want I can try and record a video of the problem and give you access to my PC if that could help.

[madshi]

I think I've reported this a long time ago but couldn't find it now. It happens almost every time I try to open a window on my primary display while using exclusive mode on the secondary. It happens both when staring a new application or just restoring a previously minimized one. It also happens when opening new windows in already running app. If you want I can try and record a video of the problem and give you access to my PC if that could help.
See my reply to Dodgexander.

[Prinz]

I just tested. I have the same "stuck window" with MPC-HC's exclusive mode for Sync Render (EVR)

So it must be a driver or OS Bug. I think it's a OS Bug, since I use a ATI card and pankov a NVIDIA card. It's unlikely that both drivers have the same bug.

Now how to get MS to fix the Bug? :(

[madshi]

Thanks for that test, Prinz, I'm really relieved that I don't have to spend time on this.

[bugmen0t]

When counting frame drops and repeats do we only want to consider if a frame is not shown exactly one time or all unwanted cases? E.g. if we play 24 frames at 48Hz then we have a pattern repeating every frame two times. So don't we want to consider everything less than two a drop and everything above two a repeat (but not counting exactly two repeats)? One could compute something like trunc(number_of_times_a_specific_frame_is_repeated - refresh_rate/framerate_of_movie) where positive values give the number of repeats and negative the drops (trunc rounds to the next integer towards zero direction). Note that for 24 frames at 60Hz this would only count the frames which are shown less than two or more than three times, so it will not detect any irregular patterns i.e. missing smoothness. That's the next thing one could think about.
Maybe someone likes to comment on this. :)

[DragonQ]

I am curious if this would create ghost image effect during the transition while motion blur effect has already existed in the 24-fps FILM-captured images...
Frame blending isn't perfect but is pretty damn good.

For example, broadcasting US sports in the UK requires the 1080i/30 (or 720p/60) video being converted to 1080i/25 but even those looking for this conversion probably wouldn't notice.

As a complete side note, ideally they'd use high-speed cameras (which they have already) to shoot at 300/600 fps, then they can downsample to both 50 fps and 60 fps without any frame trickery. This'd mean there's two "master signals" though and would probably be rather complicated to configure.

[madshi]

When counting frame drops and repeats do we only want to consider if a frame is not shown exactly one time or all unwanted cases? E.g. if we play 24 frames at 48Hz then we have a pattern repeating every frame two times. So don't we want to consider everything less than two a drop and everything above two a repeat (but not counting exactly two repeats)? One could compute something like trunc(number_of_times_a_specific_frame_is_repeated - refresh_rate/framerate_of_movie) where positive values give the number of repeats and negative the drops (trunc rounds to the next integer towards zero direction). Note that for 24 frames at 60Hz this would only count the frames which are shown less than two or more than three times, so it will not detect any irregular patterns i.e. missing smoothness. That's the next thing one could think about.
Maybe someone likes to comment on this. :)
Many things are possible, but I don't plan any of that. For the forseeable future repeated frames will only be shown if the movie framerate is roughly identical to the display refresh rate.

[Oline 61]

Hmmmm... I've changed the logic a bit in v0.85.8. In earlier versions 60i content was treated as needing 30Hz, unless DXVA deinterlacing was turned on. In that case it was treated as needing 60Hz. In v0.85.8 60i content is always treated as needing 60Hz, unless madVR knows for sure that DXVA deinterlacing won't be needed. So my best guess is that madVR asks for a 60p mode, but doesn't find it in your list. I'm not sure why it then ends up with 24p. I guess it's because 60i content can sometimes be IVTCed back to 24p. But I guess it would make more sense to use 30p in this situation. I'll put that on my to do list for the next build.

However, I'm wondering if your list of resolutions really makes sense. There is only very very few real 30p content. Most is either film content (after IVTC 24p) or video content (after DXVA deinterlacing 60p). Is your display not able to accept 1080p60?

This problem is occurring with 30p content. I don't play any interlaced content on my HTPC (most of it I ripped myself and deinterlaced/IVTCed in AviSynth before encoding with x264/xvid). As such all of my content is 24p, 25p or 30p. The specific content is from an NTSC video DVD that was deinterlaced to 30p then encoded as xvid in avi.

I've since found out that the problem is a little bit more complicated than indicated in my first post. Playing 30p x264/mkv content doesn't cause any problems but 30p xvid/avi does. I added 1920x1200p60 to the display modes and found that the 30p xvid/avi content that was switching to 24Hz now switches to 60Hz. The 30p x264/mkv content still switches to 30Hz. When playing the 30p xvid/avi content at 60Hz switching to exclusive mode triggers a resetting D3D device failed error. After a few seconds the video shows up paused and I can resume playback.

I have found I can fix the problem buy disabling deinterlacing in LAV Video. Then all 30p content plays at 30Hz. But even when deinterlacing is on auto in LAV Filters and the problem occurs, madVR reports deinterlacing off (says upstream). So not sure why changing this setting fixes it.

In any case I think the 60i logic you outlined in your response makes sense, but true 30p content should trigger a switch to 30Hz. Problem may be in LAV though.

[aufkrawall]

What does the madVR OSD say in the moment when the frame drops occur? If possible, make a screenshot of the OSD in that moment.

The drops usually occur together with frame repeats:
http://www.abload.de/thumb/fdwio9p.png (http://www.abload.de/image.php?img=fdwio9p.png)

Here is a debug log:
http://www.mediafire.com/?gso55yxhjx97vq4

[bugmen0t]

Many things are possible, but I don't plan any of that. For the forseeable future repeated frames will only be shown if the movie framerate is roughly identical to the display refresh rate.
This wasn't a request but a question about concept. What do we want to know? What should the OSD show?
I just thought about that since i was under the impression you mostly left the repeat stats for mismatching frame- and refreshrates out because of counting all repeats doesn't make much sense. So the question is what does make sense for the future?

[Prinz]

I found a workaround for the "stuck window" problem, with which I can live:

I deactivated in System - Visual Effects the animation for minimizing and maximizing of windows (don't exactly know how it's called in the english Win7 Version). Now I don't have stuck windows anymore and I can live without this animation and still use rest of Aero.

Here is the Setting that I deactivated:
http://www.addictivetips.com/windows-tips/disable-the-maximize-minimize-windows-animation-in-windows-7/