Announcement

Collapse
No announcement yet.

NVIDIA vs. Radeon VDPAU Mesa 17.2 Video Decode Performance

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

  • #31
    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


    • #32
      Originally posted by RussianNeuroMancer View Post
      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?
      They did switch to something else: nvcuvid/nvdec - access to the hardware decoders via the cuda API. Get the latest ffmpeg and mpv and do --hwdec=cuda and there you have it. Supports 10bit HEVC among other things, which VDPAU does not.

      Comment


      • #33
        cuda has some drawbacks:
        -forces GPU into P2 state, doesn't downclock
        -deinterlacing can't be activated after playback has started
        -likely no other player than mpv will support cuda

        Well, but it just works on the other hand. cuda is also best decoder on Windows (maybe with Intel d3d11va also works well, dunno).

        Comment


        • #34
          Originally posted by Gusar View Post
          They did switch to something else: nvcuvid/nvdec - access to the hardware decoders via the cuda API. Get the latest ffmpeg and mpv and do --hwdec=cuda and there you have it. Supports 10bit HEVC among other things, which VDPAU does not.
          According to Nvidia you gotta do this before you can use it in the latest mpv.....

          Using NVIDIA hardware acceleration in FFmpeg/libav requires the following steps:
          • Download the latest FFmpeg or libav source code, by cloning the corresponding GIT repositories
          • FFmpeg: https://git.FFmpeg.org/FFmpeg.git
          • Libav: https://github.com/libav/libav
          • Download and install the LATEST compatible driver from NVIDIA (Can't use the distro repository. Ubuntu Nvidia PPA has the latest driver.)
          • Downoad and install the LATEST CUDA Toolkit from NVIDIA (Gotta get it from Nvidia's website.)
          • Use the following configure command (Use correct CUDA library path in config command below)
            ./configure --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp
            --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64
          • Use following command for build: make -j 10

          Comment


          • #35
            I checked the latest FFMPEG 3.3.X build that I got from an Ubuntu update PPA in the past three days and it surprisingly doesn't have CUDA use enabled, so you have to do it manually, which explains why I felt it just wasn't working with mpv.

            Comment


            • #36
              There are no such quirks needed with a distribution that doesn't offer only outdated software.
              You just need ffmpeg 3.3 or newer and mpv 0.24 or newer in order to have cuda decoding available. In Arch you just have to install mpv and that's it.
              Last edited by aufkrawall; 05-27-2017, 05:56 PM.

              Comment


              • #37
                Originally posted by TheLexMachine View Post

                According to Nvidia you gotta do this before you can use it in the latest mpv.....

                Using NVIDIA hardware acceleration in FFmpeg/libav requires the following steps:
                • Download the latest FFmpeg or libav source code, by cloning the corresponding GIT repositories
                • FFmpeg: https://git.FFmpeg.org/FFmpeg.git
                • Libav: https://github.com/libav/libav
                • Download and install the LATEST compatible driver from NVIDIA (Can't use the distro repository. Ubuntu Nvidia PPA has the latest driver.)
                • Downoad and install the LATEST CUDA Toolkit from NVIDIA (Gotta get it from Nvidia's website.)
                • Use the following configure command (Use correct CUDA library path in config command below)
                  ./configure --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp
                  --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64
                • Use following command for build: make -j 10

                That's outdated. On a non-sucky up-to-date distro (like Arch, as aufkrawall mentioned), you need to do exactly nothing, you already have everything.

                On other distros, you need the latest ffmpeg (3.3) and simply compile it. You do not need the Nvidia SDK, the ffmpeg 3.3 release already contains the necessary stuff. It's possible you'll need the SDK for libnpp (advanced filters), but it's not needed for just cuvid. Once you have a cuvid enabled ffmpeg, compile mpv.

                The easiest way to achieve this is by using mpv-build - it's a set of scripts that compiles a local version of ffmpeg and then compiles mpv with the local ffmpeg statically linked into it, so the compiled mpv binary will not interfere with whatever outdated stuff the distro ships with.

                Comment


                • #38
                  Originally posted by Michael View Post

                  There is nothing misleading about them, qvdpautest is publicly available and has been for years, all the questions can be answered by looking at it, as well as the Phoronix Test Suite's test profile wrapped around it.
                  Have you ever thought that while pushing readers towards your other side is understandable, readers may actually give up trying to read what seems like incomplete reviews?

                  This is not a criticism, just a heads-up.

                  Comment

                  Working...
                  X