Announcement

Collapse
No announcement yet.

HD3200 and UVD2 clarification

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

  • HD3200 and UVD2 clarification

    Hello
    Just trying to sort out conflicting information on various Internet sites.

    This link says

    http://en.wikipedia.or/wiki/Unified_Video_Decoder

    "UVD2 is implemented in the Radeon RV7x0 and R7x0 series GPUs. This also includes the RS7x0 series used for the AMD 700 chipset series IGP motherboards."

    However in the immediate section below the Table lists RS780 as having UVD not UVD2.

    This pdf also seems to indicate HD3200 has UVD2

    http://www.splitted-desktop.com/stat...with_Gnash.pdf

    "This enables users to enjoy fully accelerated H.264 video playback in a web-browser on a large range of hardware including and not limited to AMD graphical processing units with UVD2 (Radeon HD 3200 IGP and Radeon HD 4000 series discrete cards)..."

    I'm just trying to figure out if I will be able to enjoy XvBA XvMC benefits ever in a future release on my 780G ...

    Cheers
    Last edited by Ragool; 12-06-2009, 03:07 AM.

  • #2
    Afaik, UVD2 can decode two H.264 streams simultaneously which the RS780 cannot.
    But UVD version is reportedly new enough to work with xvba-video.

    Comment


    • #3
      Originally posted by chithanh View Post
      Afaik, UVD2 can decode two H.264 streams simultaneously which the RS780 cannot.
      But UVD version is reportedly new enough to work with xvba-video.
      The RS780 can decode two H.264 streams simultaneously: 1 HD (1080p) and 1 SD. In practise, more streams can be supported provided the real-time constraints are respected. Otherwise, you can get frames dropped. This works equivalently to NVIDIA or Intel chips and drivers.

      Comment


      • #4
        Nvidia can decode multiply full hd videos without hacks with 195.22 driver. For older drivers you needed to export an environment var.

        Comment


        • #5
          Originally posted by Kano View Post
          Nvidia can decode multiply full hd videos without hacks with 195.22 driver. For older drivers you needed to export an environment var.
          The following applies to NVIDIA as well: "provided the real-time constraints are respected. Otherwise, you can get frames dropped". IOW, this can be slower to decode.

          Comment


          • #6
            Well i have got files with 2 video streams, but i miss a mplayer patch to switch between em on the fly - using vid option is a bit boring.

            Comment


            • #7
              Thanks for responding.

              So the 780G has UVD2? Here is more conflicting information

              "We jump from Universal Video Decoder (UVD) 1.0 on the 780G to UVD 2.0 capabilities on the 785G."

              http://www.anandtech.com/mb/showdoc.aspx?i=3615

              TIA
              Regards

              Comment


              • #8
                Originally posted by Ragool View Post
                So the 780G has UVD2? Here is more conflicting information
                This is what the driver says. IGPs are a mixture of different blocks. The VPU can be UVD2 while GPU could be older graphics core. I believe the chipset would have gained another name than a simple "update" to RS785 if more than the GPU changed.

                Comment


                • #9
                  One of the complications is that there are really about 5 different revisions of UVD out there. The versions get simplifed for marketing into "UVD1, UVD2" but that doesn't really cover all the nuances of what features & programming model are present with each class of GPU.

                  The UVD block in 780 is, IIRC, an early version of UVD2 which had most, but not all of the features we associate with "UVD2" internally. I think the official line between UVD1 and UVD2 was drawn a bit later, but from a programming standpoint the 780's decode block does fall into the UVD2 bucket.

                  New display and video features normally show up first in IGP parts, since they are expected to have a longer life in the market and since display and video are their primary purpose, rather than high performance 3D.

                  Comment


                  • #10
                    Might these nuances between video decoding blocks also explain why I am not seeing that the UVD feature is available as described here?

                    Comment


                    • #11
                      > there are really about 5 different revisions of UVD

                      > the official line between UVD1 and UVD2

                      Interesting. Obviously we should be using floating point
                      to describe this. UVD1.998 vs. UVD2.002 :-)

                      But the real question is when will we be able to decode
                      video using some variety of UVD and FLOSS-only software
                      (binary-only blobs are not acceptable)? A solution that
                      only supports UVD 3.x and newer is probably ok. Mpeg2
                      is sufficient, I don't need every obscure codec in the book.

                      The day I read about this working is the day I start
                      shopping for an ATI GPU.

                      Comment


                      • #12
                        Originally posted by Dieter View Post
                        But the real question is when will we be able to decode
                        video using some variety of UVD and FLOSS-only software
                        (binary-only blobs are not acceptable)? A solution that
                        only supports UVD 3.x and newer is probably ok. Mpeg2
                        is sufficient, I don't need every obscure codec in the book.
                        In principle never, unless somehow somebody manages to decouple the DRM protected bits from the rest--if I'm not mistaken. On top of that, apparently the OSS developers are not particularly interested on this area, be it because there isn't any monetary incentive, because of codec patents (go figure), and because you know, the majority of linux users only watch pirated movies anyway.

                        Comment


                        • #13
                          Originally posted by Dieter View Post
                          > there are really about 5 different revisions of UVD

                          > the official line between UVD1 and UVD2

                          Interesting. Obviously we should be using floating point
                          to describe this. UVD1.998 vs. UVD2.002 :-)

                          But the real question is when will we be able to decode
                          video using some variety of UVD and FLOSS-only software
                          (binary-only blobs are not acceptable)? A solution that
                          only supports UVD 3.x and newer is probably ok. Mpeg2
                          is sufficient, I don't need every obscure codec in the book.

                          The day I read about this working is the day I start
                          shopping for an ATI GPU.
                          2 points

                          first point Buy a R900 based card in the end of 2011! R900 will have opensource support for the viedeo acceleration unit...

                          second point the opensource driver for r600+/r800 will never use the UVD2 unit NEVER! but there will be a pure shader based viedeoacceleration.
                          This version will save you 50% of the needed cpu "power" (uvd2 can reduce the cpu usage up to 95%)

                          "The day I read about this working is the day I start
                          shopping for an ATI GPU."

                          really? Buy it now! the shader based version will come to the opensource driver be sure!

                          Comment


                          • #14
                            Thank you all for the useful posts esp Bridgman. I will wait for the shady er shader business to complete. When is the ETA for this? Can we expect this in Ubuntu 9.0,4 final?

                            TIA
                            Regards

                            Comment

                            Working...
                            X