• GFGJewbacca@lemm.ee
    link
    fedilink
    arrow-up
    10
    ·
    2 days ago

    I’m surprised that no one has commented on this. This is great news! It’ll be a couple of years before I build a new desktop, but I’m already looking forward to using an AMD graphics card for the first time in over a decade.

    • exuA
      link
      fedilink
      English
      arrow-up
      2
      ·
      2 days ago

      In not sure what difference it makes. AMD already supports VAAPI for accelerated video encoding/decoding.

      • frazorth@feddit.uk
        link
        fedilink
        arrow-up
        3
        arrow-down
        1
        ·
        2 days ago

        The video quality for encoding has always been bad with AMD.

        I’m hoping that by using Vulkan we can bypass poor quality encoders in drivers and get standardised accelerated encoders.

        • Laser@feddit.org
          link
          fedilink
          arrow-up
          3
          ·
          2 days ago

          From my understanding, it’s the hardware that produces bad results. There’s no encoding logic in the drivers itself. That’s why the encoding is accelerated in the first place.

          E.g. if the hardware doesn’t support b-frames – which for long it didn’t – a new driver won’t do jack. This is just about how video data gets into and out of the card, any encoding logic is handled by the hardware.

          • frazorth@feddit.uk
            link
            fedilink
            arrow-up
            2
            arrow-down
            1
            ·
            2 days ago

            If the driver is no longer using a dedicated piece of encoding hardware thats shit, but using the Vulkan logic then surely the quality would be essentially guaranteed by it being Vulkan conformant?

            The hardware wouldn’t support b-frames in this scenario, and wouldn’t matter because your just using the standard matrixes to encode the stream and if it didn’t work then surely games would also be broken.

            Or am I incorrect. Is this just standardising the API in Vulkan and it gets forwarded to the same video encoding driver? Could we not have Mesa doing a better job? 😒

            • heftig@beehaw.org
              link
              fedilink
              arrow-up
              2
              ·
              2 days ago

              It’s about adding API to Vulkan for access to the hardware encoding units that you’re complaining about.

            • Laser@feddit.org
              link
              fedilink
              arrow-up
              2
              ·
              2 days ago

              Or am I incorrect. Is this just standardising the API in Vulkan and it gets forwarded to the same video encoding driver? Could we not have Mesa doing a better job? 😒

              This is how I would read it. But if you have Mesa do it, it’s in software, and you might just be using a software coded directly then, it’s much easier.

              • frazorth@feddit.uk
                link
                fedilink
                arrow-up
                1
                ·
                1 day ago

                I wrote Mesa but meant ffmpeg. Wishful thinking that they are able to make a generic ffmpeg encoder in Vulkan to allow it to be accelerated in hardware but not relying on bad video driver codecs.

                Oh well, back to Intel.

    • heftig@beehaw.org
      link
      fedilink
      arrow-up
      3
      ·
      edit-2
      2 days ago

      Cross-platform, I guess. However, the only platform of commercial interest (and that I know of) that offers Vulkan as the primary, supported API is Android.