Announcement

Collapse
No announcement yet.

Zink Is Ending 2021 In Fantastic Shape For OpenGL Over Vulkan

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

  • Zink Is Ending 2021 In Fantastic Shape For OpenGL Over Vulkan

    Phoronix: Zink Is Ending 2021 In Fantastic Shape For OpenGL Over Vulkan

    Following all the work carried out by Mike Blumenkrantz (Valve) and others, the Mesa Zink code is ending the year in terrific and very capable shape for OpenGL running atop the Vulkan API. Here is a look at where things currently stand with mainline Mesa for Zink OpenGL-on-Vulkan compared to the native RadeonSI Gallium3D OpenGL driver.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    The result with Xonotic is encouraging. I told y'all that there would be cases where Zink+RADV would match or exceed the performance of RadeonSI, and some said no way. With some Vulkan extensions, Zink could eventually completely replace all the target-specific GL drivers for which there exists a corresponding Vulkan driver.
    Last edited by microcode; 30 November 2021, 01:23 PM.

    Comment


    • #3
      Very impressive!

      Comment


      • #4
        Zink working on Windows will certainly make AMD Windows users happy

        Comment


        • #5
          Originally posted by user1 View Post
          Zink working on Windows will certainly make AMD Windows users happy
          I'd be not so sure about "certainly", given that amdvlk basically still doesn't work at all with e.g. vkd3d-proton. They've got a bad track record of not supporting some features, extensions and not caring about bugs with API wrappers.

          Comment


          • #6
            Zink on windows would destroy the artificial crippled drivers for non workstation drivers.
            cheaper cad workstations could be possible

            Comment


            • #7
              Originally posted by aufkrawall View Post
              I'd be not so sure about "certainly", given that amdvlk basically still doesn't work at all with e.g. vkd3d-proton. They've got a bad track record of not supporting some features, extensions and not caring about bugs with API wrappers.
              AMDVLK not working with vkd3d-proton might be due to vkd3d-proton using Valve's own Vulkan extensions which were created specifically for vkd3d-proton's use case (non of which were implemented in AMDVLK). I don't think Zink uses such extensions.
              It's true that AMDVLK tends to have lots of bugs/regressions, but when it doesn't, it tends to perform pretty well. AFAIR, in the last AMDVLK test Michael did, it performed almost as good as radv in many cases, including with DXVK.
              But of course, it's up to the Zink developer to make sure it works with different drivers (I remember at first it only worked with Mesa Vulkan drivers).

              Comment


              • #8
                Originally posted by microcode View Post
                The result with Xonotic is encouraging. I told y'all that there would be cases where Zink+RADV would match or exceed the performance of RadeonSI, and some said no way.
                Yes and it wasn't even that long ago!

                Comment


                • #9
                  I see the individual OpenGL drivers being placed in a legacy path and Zink becoming the default OpenGL driver.

                  Once performance is comparable, the advantages far outweigh the negatives.

                  Comment


                  • #10
                    Originally posted by microcode View Post
                    The result with Xonotic is encouraging. I told y'all that there would be cases where Zink+RADV would match or exceed the performance of RadeonSI, and some said no way. With some Vulkan extensions, Zink could eventually completely replace all the target-specific GL drivers for which there exists a corresponding Vulkan driver.
                    Well, xonotic in super high frames per second range that happen to maybe have a low zink cpu overhead is not really necessarily representative.
                    Not saying there is no way but it's maybe too early for feeling proud of /dev/future visions.

                    Comment

                    Working...
                    X