No announcement yet.

AMD's UVD2-based XvBA Finally Does Something On Linux

  • Filter
  • Time
  • Show
Clear All
new posts

  • #76
    Originally posted by nightmorph View Post
    Meh. I'd prefer worthwhile hardware accel of regular DVDs, non-next-gen codecs (plain ol' mpeg, avi, etc.), and Flash.

    Neither AMD's Catalyst nor git checkouts of the open-source driver stack can deliver tear- and stutter-free playback of even these simple video formats. Especially not fullscreen.

    I've never used H.264 or VC-1, and I can't imagine I ever will. But I do use the other formats daily -- that's where I'd like to see some improvement. I hope this limited XvBA release is a step in that direction.
    I mostly agree with your sentiment, but you should realize that a lot of Flash video today is H.264. VP6 is still out there too of course, and obviously it'd be best if all of these codecs were well supported...


    • #77
      Originally posted by PuckPoltergeist View Post
      Seems that this get a little more complicated than I thought:
      I had that problem too.

      I got past it by linking the libtool in the libva-0.31.0 dir to my distro libtool.

      Still no luck on my ATI Radeon HD 2400 XT (on Fedora 11) though:

      libva: libva version 0.31.0-sds3
      libva: va_getDriverName() returns 0
      libva: Trying to open /usr/local/lib/va/drivers/
      libva: va_openDriver() returns 0
      vainfo: VA API version: 0.31
      vainfo: Driver version: Splitted-Desktop Systems XvBA backend for VA API - 0.5.1
      vainfo: Supported profile and entrypoints
      xvba_video: XVBA_CreateContext(): status 11
      xvba_video: XVBA_CreateContext(): status 11
      xvba_video: XVBA_CreateContext(): status 11
      X Error of failed request: BadLength (poly request too large or internal Xlib length error)
      Major opcode of failed request: 35 (X_AllowEvents)
      Serial number of failed request: 14
      Current serial number in output stream: 14


      • #78
        Originally posted by greg View Post
        The open source approach will probably not use the video decoding hardware, but shaders, and this won't be as efficient as bitstream-level acceleration -- not all decoding steps can be offloaded to hardware this way.
        thats true but... you don't get the Point...

        the Point is... the Opensource acceleration will be better because of the flexibility ...

        UVD2 can not accelerat Flac or Dirac or OGG-Theora...

        the Opensource one shader based can do this!

        in fakt Dirac is stronger/better than H264 less datarate in the same quality or better quality in the same datarate.

        UVD2 is an backward technik Obsoled the past... dirac/theora is the future and shader based acceleration is the future because of the Portabel code...

        in fakt cou can use an openCL acceleration code on nvidia and amd and intel and via in the same time! and you can use the same code on hd4000 and hd5000 and hd6000...

        und hd7000 and hd8000 and so one...

        AMD can save a lot of money by using Portabel and flexibel acceleration code! '


        • #79
          Originally posted by TeoLinuX View Post
          Yeah. Me, as a owner of a HD3650, I feel betrayed/dumped. Are UVD1 or UVD+ so different from the UVD2 that a unified or parallel driver is too much effort to implement?

          R63x is not that old!!! They're very similar to R700... why in the world are they loosing feature support?
          So disappointed

          If they hope me upgrading to R4xxx or 5xxx, well, this time I'm giving my money to nVidia. Sorry
          the fixed funktion pipeline of the UVD2 can not accelerat modern codex like the opensource Dirac oder Theora 1.1

          so in fakt the UVD2 acceleration will ever be a lame duck!

          the opensource driver for you card will have a much more flexible shader based acceleration and its posible to accelerat Dirac or theora...


          • #80
            Originally posted by Nille View Post
            Hey dont give us the infos for UVD2 the Blob work with UVD2 but it has no reasson why not on UVD
            When? what you with "older" cards? r300 up to r500?
            Older carts support focus on the opensource driver thats will be much more cost efficent for amd and for my point of view better for the linux users.

            hd2900 do not have a UVD unit 2900 has shader based viedeo acceleration.
            X1900 has shader based acceleration to.
            so yes its posible to bring portable shader based acceleration code to all card.
            older cards means hd2xxx and hd3xxx and some r500 like X1900
            i think the other cards are to old/bad...

            i think the catalyst/fglrx/UVD2 will be doomet by the opensource driver.

            first Dirac acceleration on opensource side will doom the UVD out..


            • #81
              Originally posted by droidhacker View Post
              Then I'm confused... according to this:
              HD3xxx series *does* have UVD. RHD-2xxx even.

              Or is it just fake-UVD ala shaders (aka the gallium3d plan)?
              If it is just fake-UVD, then why the distinction between R3100 (780V chipset) and RHD3200 (780G chipset)? 3100 just not strong enough? Or is it a question of being disabled in the drivers?
              HD2900 has Fake UVD there is no UVD unit.. but hd2900 is a powerfull monsterchip... this GPU can handle ALL ! be sure!
              some hd2xxx has fake chip ans some others as UVD
              all HD3000 has UVD part,, but not UVD2 part...

              but for linux H264 is patented no one want this s..t and UVD/UVD2 can not accelerat Theora/Dirac


              • #82
                Has anyone actually used this to watch a whole movie? It doesn't seem very stable to me.. I get fairly consistent mplayer hangs:

                *** glibc detected *** ./mplayer: free(): invalid next size (fast): 0x097deea8 ***
                ======= Backtrace: =========
                Just had to reboot, because killing mplayer locked up the X Server....


                • #83
                  I bought a 3850 two years ago and believed eventually my card would have hardware acceleration support. Now this day comes. But thank you ATI. You really makes me feel upset/betrayed. AMD/ATI you betray all faithful buyers who bought your RV6xx even though the performance was worse than Nvidia. I can tolerate the buggy fglrx, I can tolerate my wine broken all the time, but why fglrx can't even bring a small happiness to its RV6xx buyer?! I really think my requirement for fglrx is quite low. I just want hardware acceleration for video decoding so I can watch some 1080p video in linux environment. ATI/AMD, you really break my heart.


                  • #84

                    Did you check with mediainfo the codec used?


                    • #85
                      [mkv] Track ID 1: video (V_MPEG4/ISO/AVC)
                      [mkv] Track ID 2: audio (A_AC3)

                      It plays fine till a certain scene and then hangs mplayer.


                      • #86
                        Format/Info : Advanced Video Codec
                        Format profile : High@L5.1

                        That directly crashes here. Your info is not what i wanted to see.


                        • #87
                          What app do you use to get that? I don't have an app 'mediainfo'


                          • #88


                            • #89
                              Originally posted by doubledr View Post
                              I really think my requirement for fglrx is quite low. I just want hardware acceleration for video decoding so I can watch some 1080p video in linux environment. ATI/AMD, you really break my heart.
                              You'll get decode acceleration one way or another - might be open source drivers, might be fglrx.


                              • #90

                                Format/Info : Advanced Video Codec
                                Format profile : High@L5.1