Announcement

Collapse
No announcement yet.

NVIDIA vs. Radeon VDPAU Mesa 17.2 Video Decode Performance

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

  • #21
    I was under the impression that AMD is planning to move to VAAPI ??

    Also they promised vp9 for the latest gen of cards which still isn't there AFAIK

    Comment


    • #22
      Originally posted by debianxfce View Post
      it is easier to install custom kernel package than nv driver. You can work with desktop working and going back is easy with grub.
      Why would anyone want to use nv driver? It supports only 2D and I doubt it works with anything released in last five years.


      Comment


      • #23
        Originally posted by 89c51 View Post
        Also they promised vp9 for the latest gen of cards which still isn't there AFAIK
        Hopefully the longer they delay hardware VP9 decoding, the fewer companies will push VP9 on us.

        It's less efficient than H.264, let alone H.265.

        And as an end-user, MPEG licensing crap has nothing to do with me. Google should either just deal with the licensing or they should come up with something that is actually better before expecting hardware decoders to be made.

        Comment


        • #24
          Originally posted by 89c51 View Post
          I was under the impression that AMD is planning to move to VAAPI ??
          We support VDPAU (decode), VAAPI (encode and decode), and OpenMAX (encode and decode). Note that VDPAU does not support encode, only decode.

          Comment


          • #25
            I couldn't resist and checked how is my Radeon-equipped machine performing:

            Code:
            qvdpautest 0.5.2++
            Intel(R) Core(TM) i5-6600K CPU @ 3.50GHz
            Advanced Micro Devices, Inc. [AMD/ATI] Cape Verde LE [Radeon HD 7730/8730]
            
            VDPAU API version : 1
            VDPAU implementation : G3DVL VDPAU Driver Shared Library version 1.0
            
            MPEG DECODING (1920x1080): 121 frames/s
            MPEG DECODING (1280x720): 263 frames/s
            H264 DECODING (1920x1080): 68 frames/s
            H264 DECODING (1280x720): 159 frames/s
            VC1 DECODING (1440x1080): 186 frames/s
            MPEG4 DECODING (1920x1080): 110 frames/s
            I think it is quite apparent that the AMD cards clock only as high as they need to deliver a smooth framerate, anything more is a waste of energy. A lot more relevant test would be to force all the cards to max performance and see what they can do. As for the temperature and power consumption results, you have got to realize that comparing different chip designs, heatsink assemblies and fan speed trip points set differently by manufacturers of the specific cards being tested can be highly misleading. Also, notice that the medians of temps and watts of the Radeons appear to be at the bottom-ish end of the scale.

            Anyway, thanks for the test Michael, it was an interesting read.

            Comment


            • #26
              Is there anyway for someone to test the H.265 10-bit? I'd like to see how they preform. I'm thinking of doing a build sometime in the next year for an HTPC and I'd like to be able to support those on Linux with Kodi. I'm hoping to use AMD if possible, just because they support open source.

              Actually, is there something on Openbenchmark that might have it? I'm not great at finding stuff on there.
              Last edited by DragonDaddyBear; 26 May 2017, 04:37 PM. Reason: Openbenchmark question.

              Comment


              • #27
                Originally posted by dheitmueller View Post
                I've done a ton of testing of GPU decoding under Linux on various AMD/Nvidia platforms over the years, and there are a couple of things you may wish to consider when doing any sort of testing for VDPAU/VA-API:
                These comments would make sense... like 10 or 15 years ago. These days, if a GPU can't support 1080i decoding of any profile, in realtime, and with pretty much any filter settings, the public deserves to know.

                I agree with you that it'd be nice to know more about the content & VDPAU parameters, but only for purposes of comparison. And if you want a truly apples-to-apples comparison between Nvidia and AMD, then some trivial deinterlacing & postprocessing values should be used.

                Comment


                • #28
                  Originally posted by Holograph View Post
                  Not sure if this is pedantic or not but you have H.264 tests and you have MPEG-4 tests.

                  H.264 is part of the MPEG-4 standard... There seems to be a ton of confusion about this from a lot of reviewers that include video encoding benchmarks.

                  "MPEG-4" itself is rather inspecific. Are you talking MPEG-4 Part 2? Simple profile? Advanced simple profile?
                  Yes, you're correct about the details, but MPEG-4 part 2 is the only reasonable interpretation.

                  As for profile & level, these things also apply to MPEG-2, H.264, etc. While they'd be nice to know, what's really important are that:
                  1. The video content is realistic, meaning characteristic of what one would typically encounter "in the wild".
                  2. The same content is used to test all hardware.


                  Furthermore, as I mentioned above, filters like deinterlacing & temporal interpolation can be arbitrarily sophisticated. Especially if he's not measuring image quality (PSIM, PSNR, etc.), he should just stick to trivial values that are likely to yield nearly identical results from all implementations.

                  Comment


                  • #29
                    Originally posted by Holograph View Post

                    My reply just before yours links to a Phoronix article from 2015 that claims VDPAU does in fact support H.265 (AKA HEVC). Has that changed?
                    I suppose then that it has never worked, unless some evidence shows the opposite.
                    It certainly does not work here with a GTX 1070 and any driver, only HEVC 8 bit is available via vdpau.

                    Imho Nvidia should just put vdpau into the trash and build a wrapper for vdpau/vaapi -> cuda.

                    Comment


                    • #30
                      They seems like already put it into the thash: https://cgit.freedesktop.org/vdpau/libvdpau/log/
                      But why they doesn't switch to anything else?

                      Comment

                      Working...
                      X