Announcement

Collapse
No announcement yet.

Experimental Zink On NVIDIA's Vulkan Driver Capable Of Outperforming OpenGL Driver

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

  • #11
    Originally posted by mirmirmir View Post
    You guys are all wrong, Nvidia OpenGL sucks, and we can't confirm it because there's no proper open-source driver to compare
    Or we can try to stay objective and simply accept the fact that development of their OGL driver in terms of gaming performance has been more or less stalled, apparently in favor of Vulkan. That doesn't mean it's sucky, but some kind of jack of all trades (except for being FOSS, of course) for both legacy gaming and workstation use. It does the trick and nobody cares for OpenGL gaming anymore. Windows AMD and Intel users would still love to have this driver.

    Comment


    • #12
      Soon a world of Vulkan-only drivers and Zink.
      I am down.

      PS: To those who are like "if native GL drivers were made better than what we have, Zink wouldn't beat them anymore".
      Well, thing is that we are not living in theoreticalland, but in the real world, and so far, and Zink is on the way to beating these unperfect native GL drivers..

      Also, the positive effect of Zink is that if perf and conformance are equaling/beating native drivers, dev efforts can then shift into Zink only for GL. When dev efforts are spent on the same project, it becomes even better. Zink could in the end become the ultimate GL implementation.
      Last edited by rmfx; 18 November 2021, 02:28 PM.

      Comment


      • #13
        Originally posted by V1tol View Post
        I always said that this can be faster than native implementation.
        Faster than a bad barely maintained native implementation... if blumenkratz was working on the native driver it would almost certainly be much faster than ZINC.

        Comment


        • #14
          Originally posted by V1tol View Post
          I always said that this can be faster than native implementation.
          than slow novideo implementation. that's a low bar

          Comment


          • #15
            Originally posted by bug77 View Post

            It's always been roughly on par with their performance on Windows
            Citation needed...

            Comment


            • #16
              Take it as you want, but the fact is that zink seems to have the potential to be the fastest opengl implementation (at least) on Linux.

              Comment


              • #17
                Originally posted by cb88 View Post

                Faster than a bad barely maintained native implementation... if blumenkratz was working on the native driver it would almost certainly be much faster than ZINC.
                Gallium3d in general is a higher level API than vulkan is, there is no technical reason that a native driver would be faster than zink when all is said and done. when zink gets optimized to the brink, it could very well be faster than many vendors implementations, Vulkan drivers are supposedly much easier to write than gallium ones, and higher level ones in general, so when you have a large community optimizing drivers for gallium3d. and then you have the GPU getting much more optimized VK drivers. it could very well be that in the future, Zink will out preform native OGL.

                Comment


                • #18
                  Originally posted by Quackdoc View Post

                  Gallium3d in general is a higher level API than vulkan is, there is no technical reason that a native driver would be faster than zink when all is said and done. when zink gets optimized to the brink, it could very well be faster than many vendors implementations, Vulkan drivers are supposedly much easier to write than gallium ones, and higher level ones in general, so when you have a large community optimizing drivers for gallium3d. and then you have the GPU getting much more optimized VK drivers. it could very well be that in the future, Zink will out preform native OGL.
                  is it possible to have openGL to gallium layer just like wine-nine ?

                  Comment


                  • #19
                    OpenGL simply is old and doesn't have much incentive to get faster nowadays.

                    Even if you have such incentive, 99%+ of opengl stuff you know runs way faster then you need, meaning need to optimalizations is minimal. From perspective of Nvidia there is not much reason to upgrade them

                    Comment


                    • #20
                      Originally posted by Aryma View Post

                      is it possible to have openGL to gallium layer just like wine-nine ?
                      Im not entirely sure I understand the question. but assuming I do, Opengl already operates like that, radeonSI, Zink, freedreno, and virgl etc. are gallium drivers, the reason they work with OpenGL is because gallium3d already has a front end for opengl, (They are called state trackers).

                      Basically,

                      State tracker (Gallium9) -> Gallium3d abstraction layer -> Gallium Driver (Radeonsi) -> GPU

                      In the case of zink (State trackers can be interchangeable but do need to be manually enabled, so don't get hung up on what I use G9 works on zink too this is why d3d10 state tracker only works with llvmpipe at the moment it isn't enabled yet, and needs some work to get there)

                      Statetracker (OGL) -> Gallium3d Layer -> Gallium Driver (Zink) -> Vulkan Driver -> GPU

                      Comment

                      Working...
                      X