Announcement

Collapse
No announcement yet.

Dota 2 Radeon OpenGL vs. Vulkan Performance With Mesa Git, Linux 4.9-rc1

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

  • #21
    Originally posted by zanny View Post

    Are you just compiling your own amdgpu enabled 290 kernel or are you getting it from somewhere? I'd love to try some of these amdgpu features on my 290 but would rather not leave my PC on all night compiling the kernel for it.
    Right now I'm using Manjaro and their kernels since 4.6 have the relevant option enabled.

    You can check by running this command in a terminal: zgrep -I "CONFIG_DRM_AMDGPU_CIK" /proc/config.gz

    If the result is `CONFIG_DRM_AMDGPU_CIK=y` then your kernel should work, otherwise it may take a little more effort. What's your OS?

    Comment


    • #22
      Radv is not finished and certainly doesn't take all optimization opportunities yet.

      This is a side project by two developers that is only a couple of weeks old. That it can fully run Dota 2 and The Talos Principle (minus 4x and 8x AA) is truly amazing.

      Comment


      • #23
        ...and vulkan+Talos is faster on verde chip, they call this "semi interesting" but i agree with haagch this is amazing

        Comment


        • #24
          Originally posted by hansg View Post
          Wasn't Vulkan supposed to be faster than OpenGL? Running at half the frame rate does not do much to recommend it...
          It is faster already using both Dota and Talos, but with a blob driver.

          Here it is WIP opensource implementation, so all pieces of the puzzle are probably not there yet

          Vulkan is fine as multithread is for free, together with that adventage and if game utilize let say async compute, then voila... so, on proper VK render game done but also on proper drivers, it should go even above DX11

          Comment


          • #25
            I cant understand why everyone is testing DX11 vs DX12 or OpenGL vs Vulkan on i7-4790K with extreme resolutions!

            The main advantage of DX12 and Vulkan is the less CPU overhead. And still everyone is trying to benchmark with GPU bottleneck scenarios.

            I have an outdated phenom X4 @ 3.8GHz and switching for example from DX11 to DX12 on Tomb Raider I was able to see fantastic difference in FPS..

            Comment


            • #26
              Maybe we can ask Michael to make VK scalability testing so going down from say 960x540, then 1920x1080 and 4K... and why not even 8K, some improvise that too already

              Comment


              • #27
                The only way to see scalability on VK is to run the game in 1080p from a dual core to i7 and then do the same with OpenGL

                Comment


                • #28
                  Well no, i would start with a very bottom of it... something like Temash APU as that support Vulkan too, that will probably run great Talos lowest at @960x540

                  Comment


                  • #29
                    That is how scalability works fro me, so whatever hardware supports Vulkan (including iGPUs of course). And then 4X and again 4X and again 4X and SLI... all to the unplayable

                    Comment


                    • #30
                      From a driver POV of course as that aim to supports various hardware, that is how it should be... but if driver aim to support only one hardware, then no one needs generic Vulkan, there is GNM for that

                      Comment

                      Working...
                      X