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

[janos666]

@G_M_C

- You should use Catalyst 10.12 WHQL or later.
- If it still doesn't work, you should disable the "10-bit output". (But this version won't properly dither the "10-bit input" when "10-bit output" is disabled. However, it seems to work fine if you enable the color management function.)

Are you sure that you have a 10 bit/color compatible display and you connect it with DP, HDMI or dual-link DVI? (DVI is not recommended when DP or HDMI are available. And DP is theoretically more compatible with these VGAs in 10-bit mode.)
But you should receive an error message when your display doesn't support the 10-bit mode. So, I think the problem is an outdated display driver.


@all

By the way, I made some experiments with this build and a high quality XYZ-LUT display profile. The corrected colors with MPC-HC are very similar with the PhotoShop results when I use the absolute colorimetric intent. But PS doesn't support black point compensation with this rendering intent mode and I don't know if lcms does or it's used by MPC-HC or not... But still, I notice the highest difference on the near-black shades. And they are much darker with MPC-HC. It's a bad thing because PS would make them even lighter with black point compensation (with the relative colorimetric rendering intent where this function is available).

[Jtacdf]

Finally! revision 2840 supports bdmv mpls chapters. Unfortunately it seems that the mpeg splitter doesn't play nice with ffdshow audio decoder when switching ac3/lpcm audio streams. Internal decoder switches fine though. Anyone else got the same problem?

[Mercury_22]

@G_M_C

- You should use Catalyst 10.12 WHQL or later.
- If it still doesn't work, you should disable the "10-bit output". (But this version won't properly dither the "10-bit input" when "10-bit output" is disabled. However, it seems to work fine if you enable the color management function.)

Are you sure that you have a 10 bit/color compatible display and you connect it with DP, HDMI or dual-link DVI? (DVI is not recommended when DP or HDMI are available. And DP is theoretically more compatible with these VGAs in 10-bit mode.)
But you should receive an error message when your display doesn't support the 10-bit mode. So, I think the problem is an outdated display driver.


@all

By the way, I made some experiments with this build and a high quality XYZ-LUT display profile. The corrected colors with MPC-HC are very similar with the PhotoShop results when I use the absolute colorimetric intent. But PS doesn't support black point compensation with this rendering intent mode and I don't know if lcms does or it's used by MPC-HC or not... But still, I notice the highest difference on the near-black shades. And they are much darker with MPC-HC. It's a bad thing because PS would make them even lighter with black point compensation (with the relative colorimetric rendering intent where this function is available).

Are you using "Use application settings" in CCC's Video settings ?
If you do disable it and set Brightness and Contrast "manually" (e.g. Brightness=2, Contrast=98)
I hope this helps

[ryrynz]

The build numbers always end in .0, since the builds are incremented without ever using the last number, for the sake of tidiness wouldn't it be a good idea to remove it?

Nope.

Revision 2834 - Directory Listing
Modified Tue Jan 11 21:20:21 2011 UTC (38 hours, 6 minutes ago) by XhmikosR

-change the version scheme: MAJOR.MINOR.PATCH.X will be used, where X is the revision number
-bump version to 1.4.1

:)

[Jtacdf]

Finally! revision 2840 supports bdmv mpls chapters. Unfortunately it seems that the mpeg splitter doesn't play nice with ffdshow audio decoder when switching ac3/lpcm audio streams. Internal decoder switches fine though. Anyone else got the same problem?

Hmm. I found out that the problem is not the choice of decoder but rather lays with fast seeking just after switching streams.

[janos666]

Are you using "Use application settings" in CCC's Video settings ?

Of course, I disabled every "corrections" (khmm... abuses) in CCC. :cool:


If you do disable it and set Brightness and Contrast "manually" (e.g. Brightness=2, Contrast=98)
I hope this helps


This is something I would NOT do.
I am comparing the results of different softwares (EVR+lcms or madVR+yCMS and PhotoShop as "reference"). I would find the best solution.

[cca]

I've tried the ati 5450 at a friend's computer.

DXVA seems active but there are some bad blocking visible, especially in fade-in/outs. It seems that the decoding misses some basic "deblocking". The image isn't corrupted in any way, but in low detail zones it shows colors being too blocked than they should.

Now the odd part is that the same behavior is obtained by using CoreAVC. This decoder should do the deblocking on its own, but it is not.

Yet, I've tried with the internal FFdshow in MPC-HC, not in DXVA, and the deblocking seems working in this case.

How the hell this is possible? If it's an hardware issue then it should be shown only in DXVA. Instead CoreAVC shows it too. And I'm using latest beta for MPC-HC with the same drivers I'm running at home (but on a 4850).

With the latest builds, the renderer changes in EVR-CP seem to trigger a chroma upsampling bug in the ATI driver, NV12 no longer gives accurate video, it's blocky all over the place. Doesn't happen if YUY2 is used, but DXVA always uses NV12. I reverted to an older build and no upsampling bug. Another workaround is to use the "YV12 Chroma Upsampling" shader, but not sure how good the results are.

Using latest 10.12 Catalysts, all options in "Presentation" disabled when testing.

[janos666]

Another workaround is to use the "YV12 Chroma Upsampling" shader

Yet another is to use the 4:2:0->4:2:2 and 4:2:2->4:4:4 shaders from this (http://forum.doom9.org/showpost.php?p=1467856&postcount=170) package. They do a nice job.

[tetsuo55]

Revision 2834 - Directory Listing
Modified Tue Jan 11 21:20:21 2011 UTC (38 hours, 6 minutes ago) by XhmikosR

-change the version scheme: MAJOR.MINOR.PATCH.X will be used, where X is the revision number
-bump version to 1.4.1

:)This is the solution we wanted, now its finally possible :)

With the latest builds, the renderer changes in EVR-CP seem to trigger a chroma upsampling bug in the ATI driver, NV12 no longer gives accurate video, it's blocky all over the place. Doesn't happen if YUY2 is used, but DXVA always uses NV12. I reverted to an older build and no upsampling bug. Another workaround is to use the "YV12 Chroma Upsampling" shader, but not sure how good the results are.

Using latest 10.12 Catalysts, all options in "Presentation" disabled when testing.I believe we completely replace chroma upscaling with the shaders now, im not 100% sure though.....

[cca]

Yet another is to use the 4:2:0->4:2:2 and 4:2:2->4:4:4 shaders from this (http://forum.doom9.org/showpost.php?p=1467856&postcount=170) package. They do a nice job.

Nice catch, didn't notice the thread. Will test them.

[Fadeout]

With the latest builds, the renderer changes in EVR-CP seem to trigger a chroma upsampling bug in the ATI driver, NV12 no longer gives accurate video, it's blocky all over the place.

First, I don't understand why chroma upsampling makes the video not use deblocking. Sounds weird. What I usually noticed with lack of Chroma upsampling is that the "reds" were slightly blocky, but the effect was still very minimal, and what I see is instead major blocking going on on all the screen. Very easy to see.


I reverted to an older build and no upsampling bug.
What's the most recent that still works?

Yet another is to use the 4:2:0->4:2:2 and 4:2:2->4:4:4 shaders from this (http://forum.doom9.org/showpost.php?p=1467856&postcount=170) package. They do a nice job.

There are a bunch of them. Six of them that precisely seem to be related to do what you say.

So I need you to be more specific on what I should load.

[janos666]

So I need you to be more specific on what I should load.

4÷2÷0 to 4÷2÷2 intermediate Catmull-Rom spline5 chroma up-sampling for SD&HD video input
4÷2÷2 Catmull-Rom spline5 chroma up-sampling for SD&HD video input

But you have to use the "*on old PS 2.0 hardware" variants if you have an "ancient" VGA card, like a Geforce FX or a Radeon 9x00. :p

[Fadeout]

As I said this problem is on a 5450.

In any case, with the YV12 Chroma Upsampling default shader on the player my friend says he the problem is not fixed nor better. So I doubt it's related.

As I said it's major blocking going on here in fade-in/outs and color scales.

[G_M_C]

Thanx all :)

I'm still on catalyst 10.10, so that might be the problem. I'll update and see if that solves the problem.

[janos666]

As I said this problem is on a 5450.

In any case, with the YV12 Chroma Upsampling default shader on the player my friend says he the problem is not fixed nor better. So I doubt it's related.

As I said it's major blocking going on here in fade-in/outs and color scales.

Sorry, I missed that part with the 5450.

Catalyst 10.12 with updated DirectX and Visual C++ runtimes and those shaders with MPC-HC 1.4.1.2840 should work fine with "Full Floating Point Processing" and "Force 10-bit input" without noticeable banding or blocking.
And make sure you disable every possible image manipulation on the CCC/Video/All_Settings tab!

At least they work for me. I recently compared some images between "FFDShow software decoder + madVR" and "MPC DXVA Decoder + EVR-CP" (with those settings mentioned above).

Sorry that I am asking but... Do you use EVR-CP? Sometimes it can fall back to "System default" and use "Overlay renderer" if something is wrong with the EVR and/or decoder settings.

[cca]

Sorry, I missed that part with the 5450.

Catalyst 10.12 with updated DirectX and Visual C++ runtimes and those shaders with MPC-HC 1.4.1.2840 should work fine with "Full Floating Point Processing" and "Force 10-bit input" without noticeable banding or blocking.
And make sure you disable every possible image manipulation on the CCC/Video/All_Settings tab!

At least they work for me. I recently compared some images between "FFDShow software decoder + madVR" and "MPC DXVA Decoder + EVR-CP" (with those settings mentioned above).

Sorry that I am asking but... Do you use EVR-CP? Sometimes it can fall back to "System default" and use "Overlay renderer" if something is wrong with the EVR and/or decoder settings.

In my ATI 4850 it works fine except the chroma upsampling issue, I tested the shaders janos666 linked and it eliminated the problem. Fadeout perhaps has a different issue.

[Hera]

Observation - HalfFP doesn't stutter on ION/Atom @ 1080p @ 30FPS. Tested with CoreAVC to make certain that this is the case.

[Fadeout]

Problem fixed.

It turns out it was a drivers thing. I don't know how I fixed it but I was enabling/disabling stuff in the Catalyst Control Panel, and at some point DXVA was doing the proper deblock again. Then I went and disabled again all effects like denoise, fleshtones and dynamic contrast via DXVA checker. So now everything is ok.

THEN I tried the 422 444 shaders and they ADD very slight color blocking on the edges. The image is perfect without shaders, the image is still unchanged with the default chroma upsampling shader, but it showed the blocking when I used these new shaders.

Dunno why but *without* those shaders everything looks as it should.

[janos666]

@Fadeout

Was MLAA enabled? That thing tries to filter everything (literally). But any other "forced" AA or AF filter is a bad idea. They usually won't force anything where you want it to be forced (like in older games) but can cause problems with other applications.

You need those shaders when the default chroma resize fails. For example you need them when you enable the "Force 10-bit input" option. It's a good thing and your card if fast enough, so you should try to enable it as well as the "Full Floating Point processing".
It will minimize the remained banding.

[G_M_C]

Thanx all :)

I'm still on catalyst 10.10, so that might be the problem. I'll update and see if that solves the problem.

I've updated to 10.12 and it works. Even on my LCD on DVI :)

No clue is my HD5770 of my LCD has a dual link DVI or not, but it works, no error, no problem.

Thx peeps.

Will test now on my plasma on HDMI (connected through my receiver). Thats my secondary screen, and the one i use to watch movies (duh :p ) One question: Can i best switch the chain to xvYCC/Deep color (to accept the 10 bit) or not ?

[janos666]

I've updated to 10.12 and it works. Even on my LCD on DVI :)


No clue is my HD5770 of my LCD has a dual link DVI or not, but it works, no error, no problem.

Radeons can dither the 10-bit output for 8-bit connections (theoretically).


Can i best switch the chain to xvYCC/Deep color (to accept the 10 bit) or not ?

I don't recommend it. xvYCC is supposed to work with YCC signals but you can't get real YCC output from your VGA.

DeepColor support means that your display can receive more than 8 bits per color channels (at least 12 on paper but the standard doesn't say if it should be possible with every resolutions and if it should be visualized too or can be truncated right after the HDMI decoder chip...)
So, yes. If there is an option callled "DeepColor support" or something like that, then you should enable it.

[Mercury_22]

Of course, I disabled every "corrections" (khmm... abuses) in CCC. :cool:



This is something I would NOT do.
I am comparing the results of different softwares (EVR+lcms or madVR+yCMS and PhotoShop as "reference"). I would find the best solution.

So you're saying that using "Use application settings" in CCC's "Basic Video Color" you don't have clipping ?

Can you please tell me what settings are you using in CCC's "Color Pixel Format" (Full RGB, Limited RGB) ?
And can you test your clipping using this Test Pattern (http://sharebee.com/331f62c4)

P.S. Cause I have clipping if I don't change the default values for Brightness and Contrast in CCC's "Basic Video Color" , but I'm using HDMI and Full RGB
Also if needed you can find instructions here (http://www.videoessentials.com/dvehd/index.html)

[pankov]

Mercury_22,
in the latest Catalyst versions (since 10.8 if I remember correctly) there is a bug with many cards that Full RGB is not working over HDMI - it's the same as Limited RGB.
If your card is affected you can try using the DVI output

[Mercury_22]

Mercury_22,
in the latest Catalyst version there is a bug with many cards that Full RGB is not working over HDMI.
If your card is affected you can try using the DVI output
The clipping it's always been there for me if I use "Use application settings" in CCC's "Basic Video Color" + HDMI + Full RGB and MPC-HC (0-255 ) or WMP or WMC

[pankov]

Mercury_22,
sorry I've made a typo in my previous post - it's not "the latest version" it's "the latest versions".
Plaese, test to see if changing Pixel format makes any difference for you. If Full is the same as Limited you are affected by this bug and I think your only solution is to not use HDMI

[Mercury_22]

Mercury_22,
sorry I've made a typo in my previous post - it's not "the latest version" it's "the latest versions".
Plaese, test to see if changing Pixel format makes any difference for you. If Full is the same as Limited you are affected by this bug and I think your only solution is to not use HDMI

It always been this way for EVR (since CCC 8....) when using "Use application settings" in CCC's "Basic Video Color" for Brightness and Contrast

Tested with WMP & MPC-HC and :

1. CCC Full RGB & Limited RGB
+
2. HDMI Black Level = NORMAL & HDMI Black Level = Low

[pankov]

Sorry, but I don't quite understand your message.
Are you saying that when you change the Pixel format you get different levels on your display (without changing anything else - either settings at the display or the PC)?
If that's the case you are not affected by the CCC bug and you should be able to get the full RGB range.
Now about the clipping - it's very dependent on the decoder and rednerer that you use and of course your display settings.
Over at the madVR thread there was a discussion about this last week if I remember correctly.
In general the best configuration (quality wise) is to use RGB Full without any changes in the Basic Video Color settings and configure your display accordingly. Now if you want BTB and WTW information it's up to the renderer to not extend the video information to PC levels. In this case you should adjust your display accordingly but it's doubtful if it's quality is better than doing the conversion at the PC side. If you use madVR you can use either it's configurations or yCMS's 3DLUTs. Or if you use EVR in MPC-HC you can use lcms to get the best conversion.

[Mercury_22]

I'm saying that to avoid clipping with EVR (& CP or Sync) I have to disable "Use application settings" and use custom values for Brightness and Contrast in CCC's "Basic Video Color"

[pankov]

OK
I understand.
Sadly the quality of this conversion in CCC is not at all good. It's better to use madVR's options for levels or lcms (I'm not familiar with it but I'm sure it's able to do it) to do the conversion ... or at least this is my personal opinion.

[tetsuo55]

We have created 2 builds for testing:

Please let us know which one is faster.
You can test by "feeling/perception" and with formal benchmarks.

http://www.mediafire.com/?p5rlta1x6zkxv

EDIT: Please mention the filesize of the mpc-hc builds when reporting so the 2 versions don't get mixed up.

[burfadel]

When using Direct3D mode, there is an issue with using the next/previous buttons on a mouse. When in normal mode, the next/previous buttons (on a Logitect M950 mouse) allows you to go to the next or previous file in the folder. In Direct3D mode, those buttons simply don't work. Interestingly enough though, the next/previous buttons on the keyboard do work in Direct3D mode. Having this feature rectified would be beneficial since I'm sure there are others like myself that use the mouse as a remote control. The M950 mouse is one of those Logitech unifying devices, the range on it is quite significant.

Anyways, it would be great to have those buttons working in Direct3D mode!

[sansnom05]

Anyone who had problems described under using recent svn builds, please try this test build: http://www.mediafire.com/?rwdmy93u83q84nf

1. OSD/Subtitles rendering broken(before r2858), r2858 temporary fixed it by blur it a little, this test build should bring back sharp text once again;

2. EVR/CP cause black screen(no video), but OSD/Subtitles/Audio no problems.

:thanks:

updated test builds: http://www.mediafire.com/?d53hawr1yxobkhg

[Casshern]

Anyone who had problems described under using recent svn builds, please try this test build: http://www.mediafire.com/?rwdmy93u83q84nf

1. OSD/Subtitles rendering broken(before r2858), r2858 temporary fixed it by blur it a little, this test build should bring back sharp text once again;

2. EVR/CP cause black screen(no video), but OSD/Subtitles/Audio no problems.

:thanks:

EVR/CP still causes black screen. MPC HC x86 on Win 7 x64, ATI 5970, Catalyst 10.9. Everything up to 2808 works, 2848 does not, the test version on mediafire does not... do not know about the in between versions.

[sansnom05]

EVR/CP still causes black screen. MPC HC x86 on Win 7 x64, ATI 5970, Catalyst 10.9. Everything up to 2808 works, 2848 does not, the test version on mediafire does not... do not know about the in between versions.

If you can update your display driver to 10.12, please do that then test again, see this link: http://forum.doom9.org/showpost.php?p=1470897&postcount=15718 , one possible reason is: 10bit rgb is actually broken for recent svn builds before r2837.

[clsid]

So do I understand correctly that the black screen issue only occurs with the 10bit option enabled?

[janos666]

For me, it occured with 10-bit and 10.5<CCC_ver<10.12. And I still get black screen after exit with 10-bit + 10.12 + "Disable Aero".
But 10-bit works fine with 10.12 when I don't use the "Disable Aero" setting. And I don't think if Aero can make any difference with the D3DFS mode. So, this combination should be disabled (as a workaround).

[sansnom05]

So do I understand correctly that the black screen issue only occurs with the 10bit option enabled?

Actually it's unclear yet, cause I can't reproduce it. I just want to check this factor first.

[bobdynlan]

http://img210.imageshack.us/img210/8535/capturell.png
Add this to the list of troubles with Ati cards. Picture comes from a rd session. Ati 5750, 10.11 drv (10.12 tested without reboot, same), 7 x86, mpc-hc 2845, 10bit in, full fp, internal decoders - for xvid files. For hd (H264) the picture is perfect. Disabling 10bit in fixed the issue.
You may bash nVidia all you want, but deal-breaking issues like these are not so common.

P.S.: I don't like the very sharp OSD :)

[tetsuo55]

I just ran into that problem myself.
its caused by the "force 10bit input" option in combination with a decoder that does not support 10bit.

[Mercury_22]

The strangest thing: I've got black screen too with a VC-1 p file using internal decoder (with external decoder no black screen) with my build ( VS2010) but using XhmikosR's or Alexins's build (same revision 1.4.1.2845) with the same file and the same settings NO black screen !:confused:
I've update my VS2010 to SP1beta and recompiled the same revision and no more black screen !:confused::confused: using the same file the same settings !?

Disabling 10bit input / output or FPP had no effect and now (VS2010SP1) it's working fine with both enabled + FPP !

[janos666]

Check the About tab in their builds. They use SP1 too.

[Mercury_22]

Check the About tab in their builds. They use SP1 too.
Just Alexins it's using SP1 XhmikosR's it's not SP1 I've just checked

[Shakey_Jake33]

Does anybody else have issues with the internal MPEG2 DXVA codec (MPC version 2855)? I've been trying to play some files which are 1440x1080 interlaced files (which used to work in previous versions of MPC-HC when I last tried about 6 months ago) and they judder back and forth between frames quite badly. These same files work fine using the Cyberlink codec with DXVA enabled.

My GPU is a GeForce GT240m 1GB (driver version 266.35, GPU supporting DVXA2), OS is Windows 7 64-bit. If any other information is required, I will of course try to provide.

[namaiki]

^ Post a sample of such a video.

[Casshern]

I just ran into that problem myself.
its caused by the "force 10bit input" option in combination with a decoder that does not support 10bit.
It is not limited to any of the 10 bit options. Without any of the 10bit options i get the black screen (i never used them anyhow). It is also regardless of decoder. Newer catalysts (>10.9) have unfortunate other problems (unrelated to mpc hc, related to crossfire configurations, fan control, resolutions above 2536 horizontal), which make them unusable (at least for me). It is also unlikely that the "black screen" issue is related to the catalyst driver. My system: 980X, 5970 (which is a dual GPU board and works in crossfire mode), 12GB RAM, Windows 7 x64 professional, Cat 10.9. A friend with with an almost identical configuration (albeit cat 10.10 and a 920i7) has the exact same problem. All decoders are effected (internal and external, DXVA and non DXVA) and renderer settings do not make any difference (color management on/off, Input 10bit on/off, output 10bit on/off) etc. Also MPC HC 2808 works!

[Peekstra]

For me, the only way to get EVR-CP working is to disable full FP processing (and all other renderer settings) & DXVA and use the test build from the previous page. cat 10.12a, 7 x64, 6950

[tetsuo55]

It is not limited to any of the 10 bit options. Without any of the 10bit options i get the black screen (i never used them anyhow). It is also regardless of decoder. Newer catalysts (>10.9) have unfortunate other problems (unrelated to mpc hc, related to crossfire configurations, fan control, resolutions above 2536 horizontal), which make them unusable (at least for me). It is also unlikely that the "black screen" issue is related to the catalyst driver. My system: 980X, 5970 (which is a dual GPU board and works in crossfire mode), 12GB RAM, Windows 7 x64 professional, Cat 10.9. A friend with with an almost identical configuration (albeit cat 10.10 and a 920i7) has the exact same problem. All decoders are effected (internal and external, DXVA and non DXVA) and renderer settings do not make any difference (color management on/off, Input 10bit on/off, output 10bit on/off) etc. Also MPC HC 2808 works!My comment was directly aimed at the crushed image problem in the screenshot above my comment. The completely black screen is a different issue.
The ati drivers have recovered a lot with 10.12, but it causes crossfire problems with you.

There is still a lot of research and work being done on the renderer right now (we are lucky to have a new volunteer) so things should improve over time.
The best thing to do is open a support ticket with both your hardware manufacturer and AMD directly.
Make sure to open tickets for all the problems you experience with driver version 10.12. If you know anyone that can reproduce these problems ask them to open tickets too. AMD will put developers on a particular problem only when that problem reaches a certain quota of support tickets.

P.S the same is true for Intel and NVIDIA, dont hesitate to do the same if you have a bug on those hardwares!

[clsid]

So what is being done right now to fix the black picture problem? I think 2837 should be reverted until the cause is clear.

[tetsuo55]

This is the information i have.

issue one: Black screen on Exit when D3D fullscreen is used: This is not a regression, the renderer does not refresh the display on exit, it never has.

issue two: Crushed image (like the screenshot shown above), this has to do with the fact that some driver and codec combinations do not like "forced 10bit", again this is not a regression and will probably have to be fixed in the driver.

issue three: Black screen during playback with D3D and 10bit modes disabled, as far as i can see this is another driver bug, fixed in cat 10.12.


Conclusions:
-Ati has to fix its drivers, users need to have at least 10.12(which fixes many other bugs as well like SD dxva) (intel and nvidia seem completely unaffected).
-We need to add code to refresh the display on exit to D3D fullscreen mode.


Based on the above its easy to see that we are again running into driver bugs with ATI, and a revert is not required.


finally, nightly/unstable builds are exactly that, not meant for daily use. We would not release a stable build in this state.

[clsid]

I strongly disagree. Users should not be forced to update their drivers to fix a problem that did not occur before. That is a very bad attitude to take as a project leader. Many people don't even know what a driver is, let alone how to update one. MPC should at least provide a workaround option to get the old behavior, which might be wrong but at least works.

(talking about the black picture regression here)