• 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • 81
Initial native support for DXVA2 in SVN - Time to say goodbye to your firstborns
#46
pettson3816 Wrote:So what you're saying is that this green artifact problem will never be solved then?

Ati's technicians have been asked many times in ati's forums , by users,
about the well known limitations their drivers have. No activity seems
to be taking place , but who knows. Maybe in their next catalyst release
they will deactivate the limitation and all files will play correctly. On the
other hand maybe not. The point is that nvidia guys even use that , in
order to attack the ''other'' side.
There was a trailer from apple's site , quantum of solace in particular ,
which was used in a nvidia dxva demonstration , where the same trailer
was playing correctly using their cards , while when it was playing under
ati cards , the green artifact fenomenon was present , making the video
totally unwatcable . MPC - HC as far as i remember , has the ability to detect whether the card is able to cope with the video in dxva. That means that
if you try to play such a file using an ati card , the player simply won't use
dxva and it will revert back to software rendering .
Maybe something like that can be a part of dvdplayer also .
And don't forget. It has nothing to do with the horse-power of the cards .
All capable - dxva ati cards can decode everything , it is simply an inner
driver limitation which ati has to remove .
Reply
#47
well, well: just as I was chatting to a friend about my woes with 1080p content and no proper solution.

This is manna from heaven for my little htpc. It works beautifully even though it's only a first step.
Reply
#48
oDium Wrote:For all of you that ypu experience green artifacts in some of your files
and in quicktime trailers , you should all be aware of that this most
probably has nothing to do with xbmc itself . I bet that all of you that
having this problem , do own ati gpu's like me. Is well known that ati has
driver related dxva limitations. More specifically ati's drivers can't cope
with h.264 contect of any kind , if the reference frames in this content
is anything higher than standard bluray specs.

And nowdays there are a lot of files through the web that are encoded
beyond the standard bluray specs . All these files including full hd apple's trailers
will not play with ati cards. Well there going to play , but with dxva
off.

Apparently nvidia's cards seem to have no limitation such the above , in
their drivers. If you search the web you'll find a lot of information about the
above.

From what i've read , there is currently no intention from ati's side to
remove the limitation.

Nvidia on the other hand is even advertising the fact that it's cards can
do dxva in everything !

Yes, ATI "only" supporting H.264 up to L4.1 is obviously why 480p (L3.1) quicktime trailers, 1080p (L4.0/L4.1) quicktime trailers and unaltered Blu-Ray material (L4.1) doesn't work on either Nvidia or ATI at the moment Rolleyes.
Reply
#49
ArtVandelae Wrote:Yes, ATI "only" supporting H.264 up to L4.1 is obviously why 480p (L3.1) quicktime trailers, 1080p (L4.0/L4.1) quicktime trailers and unaltered Blu-Ray material (L4.1) doesn't work on either Nvidia or ATI at the moment Rolleyes.

To be a little more specific. From internet searching and reading a LOT of
forums and user's experiences , ati has a limitation as far as it concerns
the reference frames in a video file. Both ati and nvidia can decode in dxva
everything as far as the file remains in BLURAY original specs. If the file gets
out-of-bluray specs , then only nvidia cards will be able to decode it properly.

As far as it seems , it doesn't matter if a file has L4.1 or bigger profile , it has
to do with the reference frames. I think that ati can't cope with 5 reference
frames and above.

Some people have managed to go over the limitation using a very specific version of powerdvd
with ati catalyst 9,4 and only that !

But we all have to be aware of something else also . Not only the card's drivers have to be limited - free ,
but the player also . MPC - HC
from what i know wasn't always limited - free. It the first versions , it
also had the above bluray limitation . When the developers removed
the limitation , then again only nvidia managed to decode files with out-of-bluray specs .

Lastly i copied - paste something useful from ati forums :
'' Intel’s DXVA limitations is the same as ATI, but if these two wants to, they can follow nVidia’s lead and remove the limitation.
MPC-HC recent releases can already decode out-of-spec H.264 videos with DXVA, but that is for nVidia cards only ''
Reply
#50
First of all... well, THANK YOU to all the developers involved. XBMC devs, VLC devs, FFMPEG devs. This is a big step ahead for open source. Thank you.

Some questions:

1) Is this implementation different from CrystalHD? I have an ATI 4890 and a BCM970012 in my system (it costed little money, was curious to try it and I'm hopeful it could be the way to go to playback VC-1 interlaced stuff). Meaning, on my system, I'm gonna have a Software, CrystalHD and DXVA renderer?

2) Does this implementation open up the possibility of shaders postprocessing? MPC-HC has some really nice ones (nudge, nudge, wink, wink). Big Grin

3) In case the answer to #1 is yes (three different renderers), what's the target for the autoselection thing? Once everything will be working, what is the desired behaviour? Will this offer a "fallback mode" as MPC-HC does (where you can select a different codec for DXVA compliant material and non-DXVA compliant material)?

Once more: thank you!
Reply
#51
All I can say is OMFG Thank You! Nod

I've been watching the progress on the ffdshow tryouts SVN changelog, and I've been waiting for some time for this to come to XBMC.
Reply
#52
Aaaah, thank you!

I've bene struggling around with DSPLayer for way too long now (yet, nothing but respect to the people that develop DSPlayer) but since this is provided as a patch it maks it much easier for me to combine it with pvrtesting2, hehe Tongue Just worrying a bit about the RSTP patch that I have to use for ffdshow now.

Going to compile tomorrow and check how it does. My Sempron 140 doesn't like 1080p mkv on it's own, it will be grateful.
Reply
#53
Things i have noticed until now.

When the 'adjust refresh rate' option is off , everything seems to
be normal . Dxva is working and digital ac3 and dts passes through
my hd 4550 hdmi connection normally .

When i turn the above option on , something that is crucial for me ,
dxva option IS working , BUT a couple of things occur.

1) Refresh rate actually changes but no digital sound pass through
my hdmi . Sound only passes as analog , if not at all.
When i set my sound settings to analog only , everything is ok.

2) When a file stops playing by itself and not by pressing stop on the
keyboard , xbmc doesn't show up. It only shows a black screen , but
it is actually running . When i hit alt+f4 , it closes and i go to windows
desktop normally. When i hit alt+ctrl+del and i see the task manager
xbmc is still running . Just saying this to clear that is doesn't show as
''not responding'' or something.

That's all . Keep up the good work guys !!
Reply
#54
Win 7 x64, Ati 5750 and just black screen when using DXVA (with sound, GUI and subtitles). Whats wrong?

Some message appears on screen when I press play, but its so fast, that i am not able to read it...
Reply
#55
MartijnGP Wrote:Aaaah, thank you!

I've bene struggling around with DSPLayer for way too long now (yet, nothing but respect to the people that develop DSPlayer) but since this is provided as a patch it maks it much easier for me to combine it with pvrtesting2, hehe Tongue Just worrying a bit about the RSTP patch that I have to use for ffdshow now.

Going to compile tomorrow and check how it does. My Sempron 140 doesn't like 1080p mkv on it's own, it will be grateful.

RSTP patch?

K
LG OLED55BX6LB, Zidoo Z1000 Pro (ZDMC 20.2 Arctic Zephyr 2 Resurrection), Yamaha RX-A3060, Polk Signature Fronts & Centre, Wharfedale D300 Atmos surrounds, Heco Victa Prime Sub 252A, DSPeaker Antimode 8033 Cinema
Reply
#56
elupus Wrote:Just a note.. The video scalers are doing nothing for dxva at the moment. I render directly to backbuffer so it's all how dxva processor scales at the moment.

And as i noted, it's still a WIP, untill i get some sort of detection working as to what will work and what will not it can't really be used auto.

Elis:
Any reason you say it's better to render to a texture then to backbuffer? we don't do any 3d stuff, and that would just be another step of renderering + to use any of the more advanced dxva processors, you'd have to render to a texture of backbuffer size.

It would allow the use of shaders to continue processing the video i suppose. But i'm not sure how big benefit that would be.

The late loading I would really like to avoid, but i can be convinced otherwise if it will fail totally on XP (which i suppose is likely now that i think of it). It's not that many functions that needs to be getproced.

I can think of 2 reasons why to use textures and not the backbuffer:
- Its easier to manipulate the video. Sometime you may want to render the video in a small preview window, or render somthing beneath it. It makes your life a lot easier since all of this is allready in the framework.
- Intel cards don't work well with that (if at all). They usually require a POT render target to work

The performance difference is not significant. (I checked that)

As far as the late DLL loading, you only need to get the address of 2 functions - 1 for the decoding and and 1 for VP. This is a must since XP version will not work otherwise.

Some comments regarding the green screens and green pixel:
- I've tried a couple of H.264 movies, and it looks like the VLC decoder is not handling everything correctly. There are alot of patches from the past from both FFDShow and MPC that enable extra features not corrently supported by the "official" FFmpeg.

- ATI is corrently limited by the number of refrence frames. I think the last catalyst driver fixed that.
Reply
#57
oDium Wrote:To be a little more specific. From internet searching and reading a LOT of
forums and user's experiences , ati has a limitation as far as it concerns
the reference frames in a video file. Both ati and nvidia can decode in dxva
everything as far as the file remains in BLURAY original specs. If the file gets
out-of-bluray specs , then only nvidia cards will be able to decode it properly.

As far as it seems , it doesn't matter if a file has L4.1 or bigger profile , it has
to do with the reference frames. I think that ati can't cope with 5 reference
frames and above.

Again, the videos currently giving the DXVA decoder problems are not out-of-spec clips that push decoders to their limits, but standard blu-ray compliant clips that play on everything including a Blu-Ray player, a Broadcom card, the PS3/360 and any other consumer devices that do hardware decoding. Trying to play a video with too many reference frames on an ATI card (like the Quantum of Solace trailer) just gives you garbled, yet visible, video, not the solid green screen that many clips show in XBMC on both Nvidia and ATI cards.
Reply
#58
Elis - Running an ATI HD4670 - just upgraded to 10/1 drivers, but the ref frames issue still seems to be prevalent. All failed movies have high reference frames (5 and over).
Reply
#59
I just conceived a child for the sole purpose of giving it to elupus when it is born.
Reply
#60
oldpainless Wrote:RSTP patch?

K

* RTSP, excuse me.

I mean the one mentioned in this thread:
http://forum.xbmc.org/showthread.php?tid=65488
It's needed to be able to play the RTSP stream from MPTV in DVDPlayer.
Reply
  • 1
  • 2
  • 3
  • 4(current)
  • 5
  • 6
  • 81

Logout Mark Read Team Forum Stats Members Help
Initial native support for DXVA2 in SVN - Time to say goodbye to your firstborns2