Zink OpenGL-On-Vulkan vs. RadeonSI OpenGL Performance As Of January 2021

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • phoronix
    Administrator
    • Jan 2007
    • 67391

    Zink OpenGL-On-Vulkan vs. RadeonSI OpenGL Performance As Of January 2021

    Phoronix: Zink OpenGL-On-Vulkan vs. RadeonSI OpenGL Performance As Of January 2021

    With the Zink OpenGL-on-Vulkan implementation within Mesa on a nice upward trajectory with most recently now having the backing of a Valve contract developer and a focus on getting the backlog of patches to this Gallium3D code upstreamed, here are some fresh benchmarks looking at where the performance currently stands when using Zink atop the RADV Vulkan driver compared to using the native RadeonSI driver with this round of testing from a Radeon RX 5700 XT graphics card.

    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
  • kbios
    Senior Member
    • Dec 2009
    • 277

    #2
    Very interesting! I suppose the reason for the performance difference is that zink is more tuned for the intel vulkan driver?

    Comment

    • TheOne
      Senior Member
      • Mar 2013
      • 469

      #3
      Seems like a long way to go on the performance department.

      Comment

      • schmidtbag
        Senior Member
        • Dec 2010
        • 6618

        #4
        I expected it to be slower than native, but not that much. Seems something is off.

        Comment

        • Snaipersky
          Senior Member
          • Mar 2016
          • 285

          #5
          There's a lot of performance work still sitting in mike's repo vs mainline, like resource caching (if I recall he said it was a double digits improvement on AMD, and other work.

          Comment

          • oibaf
            Senior Member
            • Feb 2008
            • 1223

            #6
            Originally posted by kbios View Post
            Very interesting! I suppose the reason for the performance difference is that zink is more tuned for the intel vulkan driver?
            I myself tested it and I found a similar performance ratio between zink and intel.

            Comment

            • QwertyChouskie
              Senior Member
              • Nov 2017
              • 638

              #7
              Interesting, but would have rather seen the zink-wip branch benchmarked

              Comment

              • dokipen
                Junior Member
                • Jan 2021
                • 4

                #8
                I expect that stability and correctness are the main goals of getting the patches into main. Then Mike's performance work can trickle in and well see improvements over the year ahead.

                Comment

                • baryluk
                  Senior Member
                  • Jul 2010
                  • 271

                  #9
                  Michael Agreed. You should check out zink-wip branch.

                  I run zink daily, and get 75-90% of radeonsi on my GPU.

                  I do run Unigine Heaven, Unigine Valley, Unigine Superposition, Supertuxkart, glmark2 and GpuTest 0.7 almost daily for performance, and your results are nothing like mine. I get easily 65-75% in all these benchmarks compared to radeonsi on same GPU.

                  Also check Unigine Superposition. It works perfectly with zink-wip, and often 80-90%, depending on resolution and other settings.
                  Last edited by baryluk; 23 January 2021, 11:29 AM.

                  Comment

                  • baryluk
                    Senior Member
                    • Jul 2010
                    • 271

                    #10
                    Originally posted by kbios View Post
                    Very interesting! I suppose the reason for the performance difference is that zink is more tuned for the intel vulkan driver?
                    Not really anymore. I was helping Mike for some time with AMD/Radv performance, and for few weeks he also has own AMD GPU to test. We can get easily 70%-85% of radeonsi in most games, benchmarks and microbenchmarks. But that is with proper branches, flags and compiler options, etc.

                    Comment

                    Working...
                    X