Media Player Classic Home Cinema (MPC-HC) - DXVA!

[jq963152]

Originally Posted by nevcairiel

4K decoding works with CUVID on my 680.
I'll look into DXVA soon, there were some driver bugs that caused it to fail in earlier versions, but maybe those are fixed now.

Thats because the 520 does not have enough memory bandwidth for "copy back" modes like CUVID or DXVA2-CB, its to be expected, the card is too weak.
Although the decoder chip isn't really all that fast on 4K content either, it can do 24p just fine, but above 30fps it'll start to lag.

And that's why it would probably be appreciated if there would be a way to be able to test if it would work with DXVA2 (native). And to be able to test if there would be less lag with DXVA2 (native) than with CUVID and DXVA2 (copy-back) .

And that's why it was requested if it would be possible to add a "4K checkbox" (or whatever you wanna call it) to LAV Video Configuration panel and a "Skip framesize check" (or whatever you wanna call it) setting to MPC-HC DXVA Settings .

At least as long as auto-detection is not working properly .

[jq963152]

PS:

Originally Posted by nevcairiel

Although the decoder chip isn't really all that fast on 4K content either, it can do 24p just fine, but above 30fps it'll start to lag.

Well, at least according to the following Wikipedia page:

http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC#Levels

AVC Level 5.1 appears to be limited at 30 fps with "4K", doesn't it?

60 fps with "4K" would be AVC Level 5.2, wouldn't it?

So, wouldn't what you wrote above maybe "just" mean that it probably can do Level 5.1 but not Level 5.2?

[v0lt]

@jq963152

But what about what Wikipedia says for example:

What does it change?

How did you test that if it is currently blocked in MPC-HC?

I used a special test builds.
http://www.mediafire.com/?n6ajswiwd3q7h (use at your own risk)

AVC HP@L5.1 apparently would allow for much higher resolutions than 1920x1080, wouldn't it?

What is the question? Show specification Nvidia, which states full support for AVC HP@L5.1.
We're talking about the real possibilities of video cards.

Is this:

http://us.download.nvidia.com/XFree8...limits-decoder

your source for what you just posted?

Yes. This is the official source. I saw it.

VDPAU Feature Sets C and D
VDP_DECODER_PROFILE_H264_MAIN, VDP_DECODER_PROFILE_H264_HIGH:

Complete acceleration.

Minimum width or height: 3 macroblocks (48 pixels).

Maximum width or height: 128 macroblocks (2048 pixels).

Maximum macroblocks: 8192

I see no support for 4K.
But this information is also incorrect. In reality, there must be so:

VDPAU Feature Sets C and D
Maximum width or height: 128 macroblocks (2032 pixels).

Maximum macroblocks: 8190

If you need a blank screen, then you can use the Microsoft DTV-DVD Video Decoder.

[jq963152]

Originally Posted by v0lt

I used a special test builds.

Are you a developer of MPC-HC? Is there any "official" link where this would be available?

Originally Posted by v0lt

I see no support for 4K.
But this information is also incorrect. In reality, there must be so:

VDPAU Feature Sets C and D
Maximum width or height: 128 macroblocks (2032 pixels).

Maximum macroblocks: 8190

Then what about the following article for example:

Originally Posted by anandtech.com/show/4380/discrete-htpc-gpus-shootout/11

http://www.anandtech.com/show/4380/d...us-shootout/11

[...]

The GT 520's scores above are more interesting. Even the high end GPUs such as the 460 and 560 are unable to achieve that frame rate. The answer was buried in the README for the latest Linux drivers. The GT 520 is the first (and only GPU as of now) to support the VDPAU Feature Set D.

We asked NVIDIA about the changes in the new VDPAU feature set and what it meant for Windows users. They indicated that the new VPU was a faster version, also capable of decoding 4K x 2K videos. This means that the existing dual stream acceleration for 1080p videos has now been bumped up to quad stream acceleration.

[...]

?

Originally Posted by v0lt

If you need a blank screen, then you can use the Microsoft DTV-DVD Video Decoder.

Just tried "Microsoft DTV-DVD Video Decoder" with MPC-HC x64 1.6.3.5818 here with GT 520 (GF119) and 3840x2160p and 4096x2304p testclip and apparently the screen remains blank (black) indeed, apparently just audio is being played back and GPU-Z/NVIDIA Inspector for example apparently do not show any VPU Load at all.

But would that automatically "translate" to MPC-HC Video Decoder and LAV DXVA2 (native) ?

And, again, see for example:

Originally Posted by jq963152

Using GT 520 (GF119) here with LAV Filters 0.51.3, both, LAV CUVID and LAV DXVA2 (copy-back) apparently are shown as being "active" in the LAV Video Configuration panel when playing back 3840x2160p testclip and even when playing back 4096x2304 testclip. But, as mentioned in the quote above, it appears to be very laggy with LAV DXVA2 (copy-back). With LAV CUVID it appears to be much better, although, unfortunately, not necessarily perfect either.

And even "nevcairiel" wrote the following:

Originally Posted by nevcairiel

4K decoding works with CUVID on my 680.
I'll look into DXVA soon, there were some driver bugs that caused it to fail in earlier versions, but maybe those are fixed now.

Thats because the 520 does not have enough memory bandwidth for "copy back" modes like CUVID or DXVA2-CB, its to be expected, the card is too weak.
Although the decoder chip isn't really all that fast on 4K content either, it can do 24p just fine, but above 30fps it'll start to lag.

So what about MPC-HC Video Decoder and LAV DXVA2 (native) ?

[DragonQ]

How can I tell if MPC-HC is doing IVTC correctly? I'm using LAV Filters and EVR. The source is mostly 24p but with a few 60i bits, so most of it needs to be ITVCed. When I step frame-by-frame it looks like it's just playing it back at 60p and not IVTCing correctly, but I think it looks OK when played back at full speed. However, the frame rate indicator (CTRL+J) always says 59.xxx FPS, never 23.976 FPS, so maybe it's not doing IVTC properly?

[nevcairiel]

EVR* does not perform full removal of hard-telecine. It will detect the cadence, and ensure the frames are properly stitched together, but it will not remove duplicate frames, which results in 60p output - artifact free, but not decimated to 24p.

* Technically thats not EVRs doing, but the GPU drivers

[madshi]

Originally Posted by DragonQ

How can I tell if MPC-HC is doing IVTC correctly? I'm using LAV Filters and EVR. The source is mostly 24p but with a few 60i bits, so most of it needs to be ITVCed. When I step frame-by-frame it looks like it's just playing it back at 60p and not IVTCing correctly, but I think it looks OK when played back at full speed.

You'll get the typical 3:2 motion judder this way. You can play such videos at proper 24p by either using the DScaler IVTC Mod MPEG2 Decoder (works only for MPEG2, obviously), or by using some funny AviSynth scripts, or by using madVR's built in IVTC algorithm.

[DragonQ]

I see. So if I wanted to play these files using MediaPortal (which only supports EVR), I'd have to set the MPEG2 decoder to DScaler IVTC Mod MPEG2 Decoder?

[madshi]

Yep...

[DragonQ]

OK. I can't seem to get good playback in MPC-HC with MadVR though. The only way I can get the stats to say "ivtc" is by ticking "disable automatic source type detection" and choosing "force film mode". However, I then get juddery video and combing artefacts.

[madshi]

Originally Posted by DragonQ

OK. I can't seem to get good playback in MPC-HC with MadVR though. The only way I can get the stats to say "ivtc" is by ticking "disable automatic source type detection" and choosing "force film mode". However, I then get juddery video and combing artefacts.

Do you get juddery video and combing artifacts with movie content or with video content? Of course forcing film mode won't work for anything other than native film content.

[DragonQ]

I'm pretty sure it's 24p material that has been telecined to 60i. If I force IVTC in MadVR and step through the frames, no frames are repeated (as expected) but there's still combing artefacts on some of them. Here's a sample.

[madshi]

From a quick check the sample doesn't seem to be 24p -> 60i content. It's something else. IVTC is not possible with this sample. Consequently DScaler2 IVTC Mod shows the same combing as madVR's IVTC algorithm.

[DragonQ]

Originally Posted by madshi

From a quick check the sample doesn't seem to be 24p -> 60i content. It's something else. IVTC is not possible with this sample. Consequently DScaler2 IVTC Mod shows the same combing as madVR's IVTC algorithm.

Any ideas what it could be then?

[madshi]

Would have to look at every field separately to say for sure. Probably it was originally movie content, but then it got half converted to video mode, with different fields being blended into each other. This is sometimes done to bring 25p content to 24p or vice versa.

[DragonQ]

Ah that could be it...the show is a mixture of US and Japanese footage, so it should be 24p all the way through. However, I know this particular episode was sourced from an "international master", so it could be a weird NTSC -> PAL -> NTSC conversion, which I imagine would be a major pain to get proper playback with.

I'll check some other episodes tonight that weren't from international masters and see if I get the same result.

[DragonQ]

Hmm, other episodes are the same. I guess the best way to play it back is deinterlaced to 60p but that still produces combing and it's a bit weird that IVTC doesn't work when 60p produces a frame repeating pattern of 3-2-3-2-3-2...

What program allows one to look at each field separately?

[madshi]

Originally Posted by DragonQ

Hmm, other episodes are the same. I guess the best way to play it back is deinterlaced to 60p but that still produces combing and it's a bit weird that IVTC doesn't work when 60p produces a frame repeating pattern of 3-2-3-2-3-2...

My AMD card's DXVA seems to try to IVTC this sample, too, with combing results. I think AMD DXVA is mistaken, trying to IVTC this. In theory it's supposed to automatically detect whether IVTC or video mode deinterlacing should be performed. This detection seems to fail with this sample. However, if you enable "agressive deinterlacing" + "force deinterlacing" + "enable YADIF deinterlacing" in LAV, the combing is gone.

Originally Posted by DragonQ

What program allows one to look at each field separately?

AviSynth. Or a special madVR debug build (not polished enough to be published).

[ikarad]

With some blu-ray movies, some subtitles are not displayed with mpc-hc 1.6.4.5957 (this problem exist since MPC-HC 1160)
(this problem exists from the start of HD subtitle support).

When two subtitles must be displayed in the same time, only one subtitle is
displayed.

file

http://www.mediafire.com/download.php?2fwqy0doid91uj1
http://www.mediafire.com/download.php?tsr8zol53hvdt9s
http://www.mediafire.com/download.php?9se81h31ciggp4b

bugtraq
https://sourceforge.net/apps/trac/mpc-hc/ticket/48#comment:28

[Xaurus]

Originally Posted by nevcairiel

There is not.
Not to mention that a decoder alone wouldn't be enough, you also need a renderer that can render 3D signals, which also doesn't exist yet.

So if no decoder exists what is this?

http://corecodec.com/products/coremvc

Or is this something else? I have a 3D TV which I would like to use with my htpc...

As you say.. won't work without a compatible renderer but that Stereoscopic player must have a renderer of sorts?