Announcement

Collapse
No announcement yet.

Zink OpenGL-On-Vulkan Benchmarks Are Very Competitive To Radeon OpenGL Driver

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

  • Zink OpenGL-On-Vulkan Benchmarks Are Very Competitive To Radeon OpenGL Driver

    Phoronix: Zink OpenGL-On-Vulkan Benchmarks Are Very Competitive To Radeon OpenGL Driver

    With this weekend having seen more Zink refactoring code land and Zink being faster than RadeonSI at least for some operations, it was time to fire up some fresh benchmarks of this generic OpenGL-on-Vulkan implementation. From the newest Mesa code this weekend after the latest Zink patches were merged, here is a look at how the Zink performance is compared to the RadeonSI Gallium3D driver's native OpenGL support. All of the testing was done using an AMD Radeon RX 6800 XT graphics card.

    https://www.phoronix.com/review/zink-radeon-august2022

  • #2
    I think when RadeonSi will work with ACO, the gap between Zink and RadeonSi will widen again.

    Comment


    • #3
      Thanks for those tests!
      Strange how in some games the low resolution falls far behind but at 4K Zink gets 30% ahead. Overall it's allready much better than I thought.

      Comment


      • #4
        Have Zink reached its full potential or is there still room for further improvement?

        Comment


        • #5
          Does it works over nvidia proprietary driver? I tried to compile it and run on my N Switch but could not make it work...

          Comment


          • #6
            When taking the geometric mean of all the benchmarks carried out for this article, Zink was running at 93% the speed of the RadeonSI driver for the wide range of games/tests! Overall this was the most impressive showing we have seen out of Zink to date and for a number of games this open-source OpenGL-on-Vulkan implementation either patched or outperformed AMD's RadeonSI OpenGL driver. Especially with OpenGL use tending to be now for older Linux games with more of the newer and heavier titles focusing on the Vulkan API directly, even in cases where Zink didn't outrun RadeonSI it still tended to deliver a very sufficient experience for a current-generation graphics card.
            I'm not really sure that we can extrapolate the results and the overall fitness of Zink to handle OpenGL in a current high end card like the Radeon 6800 and an high end cpu like the 12900K (specially when for some of these are getting fps > 900).
            Michael, any plans to re-do these tests with something more inline with mainstream graphics cards and cpus from your hardware stable (something like the 6600 and i5, or ryzen x600)?

            Comment


            • #7
              Originally posted by uid313 View Post
              Have Zink reached its full potential or is there still room for further improvement?
              Radv and zink are young and I am sure there are plenty of possible optimizations can be done.

              Comment


              • #8
                Originally posted by phoronix View Post
                trusty 6800 XT
                They are FINALLY becoming affordable... and are already in the "classic, trusty ol' workhorse" category of cards. Time flies I guess.

                Originally posted by user1 View Post
                I think when RadeonSi will work with ACO, the gap between Zink and RadeonSi will widen again.
                Honestly for some reason I thought ACO was the default since Mesa 20.x -- not sure why I believed that to be the case!

                Originally posted by Anux View Post
                Strange how in some games the low resolution falls far behind but at 4K Zink gets 30% ahead. Overall it's allready much better than I thought.
                I'm curious to know what this indicates! Could it be that there is more CPU overhead at lower resolutions with Zink but at higher resolutions during GPU-bound activities it starts to shine? If so, how? Because in that case... ... then it seems like the answer to uid313 's question lies within the answer re: better perf @higher resolutions vs @lower resolutions.

                -----

                I'm curious to see if Zink works on semi-supported Vulkan implementations like with Intel's Ivy Bridge and Haswell iGPUs... and then also wonder how it performs on Southern Islands and Sea Islands cards.

                Comment


                • #9
                  Originally posted by V1tol View Post
                  Radv and zink are young and I am sure there are plenty of possible optimizations can be done.
                  I wouldn't call Radv "young", it exists since 2016, the year when Vulkan itself was released and it's already very well optimized.

                  Comment


                  • #10
                    Originally posted by Eirikr1848 View Post
                    Honestly for some reason I thought ACO was the default since Mesa 20.x -- not sure why I believed that to be the case!
                    It is the default since Mesa 20.2, but only with Radv. ACO support for RadeonSi is still WIP.

                    Comment

                    Working...
                    X