Announcement

Collapse
No announcement yet.

State of video playback on ATI

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • State of video playback on ATI

    Been trying to find a definitive answer, but as a relative n00b I guess I just don't know where to look.

    I just got a new ASUS M3A78-EM mobo (780g chipset, onboard HD3200 video) for my brand new HTPC/MythTv box.

    And what I'd like to know is, what is the state of hardware video decode acceleration (for the various video formats such as MPEG2 and H.264), for either the open source or proprietary drivers, for this model of video card?

    (And the follow-up question, implied: Do I go back to the store, where I actually ordered an M3N78-EM (Nvidia), and get them to swap the one I got for the one I asked for?)

    Big thanks in advance.

  • #2
    Originally posted by Rodyland View Post
    Do I go back to the store, where I actually ordered an M3N78-EM (Nvidia), and get them to swap the one I got for the one I asked for?)

    Big thanks in advance.
    Simple answer: yes.

    Currently linux support for that chip is lacking and nvidia has this.

    Comment


    • #3
      Originally posted by Rodyland View Post
      Been trying to find a definitive answer, but as a relative n00b I guess I just don't know where to look.

      I just got a new ASUS M3A78-EM mobo (780g chipset, onboard HD3200 video) for my brand new HTPC/MythTv box.

      And what I'd like to know is, what is the state of hardware video decode acceleration (for the various video formats such as MPEG2 and H.264), for either the open source or proprietary drivers, for this model of video card?

      (And the follow-up question, implied: Do I go back to the store, where I actually ordered an M3N78-EM (Nvidia), and get them to swap the one I got for the one I asked for?)

      Big thanks in advance.
      The open source drivers have video rendering (Xv) support (colorspace conversion and scaling), but no decode support yet. However, we have released the 3D engine documentation, so it should be possible to implement this, it's just not done yet.

      Comment


      • #4
        Originally posted by Rodyland View Post
        .....
        If you want now ! Video Decode choose an Nvidia Card if you can wait and free drivers are important for you choose the ATI

        Comment


        • #5
          Originally posted by Nille View Post
          If you want now ! Video Decode choose an Nvidia Card if you can wait and free drivers are important for you choose the ATI
          I can wait... I can't actually use the inbuilt graphics yet because my TV is composite only... So until I upgrade my TV to something digital I'll be using a spare PCIe video card I have.

          But the question is, how long is the wait expected to be, by those in the know? 1 month? 6 months? 2 years? 1-6 months I can live with quite easily.

          Comment


          • #6
            Get a cheap nvidia 8+ card with 512 mb vram - that works now and what is in a year does not really matter. Then you can buy the next card

            Comment


            • #7
              Originally posted by Kano View Post
              Get a cheap nvidia 8+ card with 512 mb vram - that works now and what is in a year does not really matter. Then you can buy the next card

              I think that's what I've decided to do. I got an nvidia 9x (or something) from a friend who didn't need it. It's got no fan so it'll be nice and quiet, too.

              Comment


              • #8
                Originally posted by Rodyland View Post
                Been trying to find a definitive answer, but as a relative n00b I guess I just don't know where to look.

                I just got a new ASUS M3A78-EM mobo (780g chipset, onboard HD3200 video) for my brand new HTPC/MythTv box.

                And what I'd like to know is, what is the state of hardware video decode acceleration (for the various video formats such as MPEG2 and H.264), for either the open source or proprietary drivers, for this model of video card?

                (And the follow-up question, implied: Do I go back to the store, where I actually ordered an M3N78-EM (Nvidia), and get them to swap the one I got for the one I asked for?)

                Big thanks in advance.
                I'd go with AMD each way you look at it. The radeon driver works like a charm (and I'm using a fairly old version), und ffmpeg-mt can act as fine bridge until hardware video decode acceleration kicks in.

                Comment


                • #9
                  Has there been any word on how much video decoding is being planned for the r600 and upwards chipsets? I'm the proud owner of an RV670 with a Q6600 but struggling to watch 1080p h.264 encoded video under GNU/Linux.

                  Also, will video decoding be part of the radeon / radeonhd driver, or is that outside the scope of these projects?

                  -JBM

                  Comment


                  • #10
                    Quick answer is "we're not sure yet, but there will be more than there is today". I see three parts to the solution :

                    1. A "bottom up" implementation in the open source drivers using shaders, starting with MC and filtering (the most expensive parts of the decode stack) and moving up as far as seems to make sense. We may start with XvMC as an example for other codecs.

                    2. An *attempt* to allow the use of UVD/UVD2 in open source drivers. We have committed to try, but I don't know if this will work or not. The primary risks here all relate to making sure we don't weaken the robustness of DRM implementations on other OSes by providing information to support Linux.

                    3. Implementation of decode acceleration in the fglrx driver.

                    The question of "where does the video decoding code go ?" is a tricky one. The Xv acceleration code is in the X driver (radeon/radeonhd) and is used "indirectly", ie through the X wire protocol, but I'm told that even the existing XvMC code in X drivers is actually "direct rendering" code (like Mesa) but was put into the X driver source tree for convenience.

                    XvMC seems to be defined primarily for "indirect rendering" (although I hear conflicting views on this), while VDPAU seems to be defined for direct rendering only and VA-API seems to be defined for both.

                    So... bottom line is that we're not quite sure where the final open source video decode logic will go yet
                    Last edited by bridgman; 04-22-2009, 02:39 PM.

                    Comment


                    • #11
                      Originally posted by bridgman View Post
                      2. An *attempt* to allow the use of UVD/UVD2 in open source drivers. We have committed to try, but I don't know if this will work or not. The primary risks here all relate to making sure we don't weaken the robustness of DRM implementations on other OSes by providing information to support Linux.
                      With other OSs, you surely mean Windows, right? I believe the radeon driver is easily portable to the *BSDs etc. I might be wrong, but I believed that was one of its positive aspects.

                      Comment


                      • #12
                        Without getting too specific, think about which popular OSes do *not* use the X windowing system and *do* require robust DRM.

                        In other words, not BSD or Solaris
                        Last edited by bridgman; 04-22-2009, 03:26 PM.

                        Comment


                        • #13
                          So... what bridgman is saying is that the DRM requirments of "other OSes" is making driver implementation on linux harder?

                          Gee, what a surprise! Who could have seen that one coming?!

                          Comment


                          • #14
                            Originally posted by jimbobsma View Post
                            Has there been any word on how much video decoding is being planned for the r600 and upwards chipsets? I'm the proud owner of an RV670 with a Q6600 but struggling to watch 1080p h.264 encoded video under GNU/Linux.

                            Also, will video decoding be part of the radeon / radeonhd driver, or is that outside the scope of these projects?

                            -JBM
                            If you use mplayer (or smplayer, anything that uses mplayer as the actual video player), the latest versions of mplayer-mt have worked wonders for me (the bit about the git pulls and using the ffmpeg-mt branch): http://www.mplayerhq.hu/design7/news.html

                            I have an Athlon X2 7750 Black edition that your Q6600 is almost certainly faster than and the multithreaded mplayer + the flicker-free xv video in the open-source ATI driver makes my 3200HD great for 1080p in my media center. No more stuttering or struggling with videos. I'd rather have the UVD working so I could use a much slower CPU that generates less heat, but since you have the CPU power, may as well fully use it. I don't think any distros ship with it by default, so you'll have to compile it yourself, but it's pretty stable and works well.
                            Last edited by Pitabred; 04-22-2009, 05:51 PM.

                            Comment


                            • #15
                              Originally posted by bridgman View Post
                              2. An *attempt* to allow the use of UVD/UVD2 in open source drivers. We have committed to try, but I don't know if this will work or not. The primary risks here all relate to making sure we don't weaken the robustness of DRM implementations on other OSes by providing information to support Linux.
                              Gee Hollywood, thanks a lot .

                              Anyway, your efforts are appreciated bridgman, cheers .

                              Comment

                              Working...
                              X