Announcement

Collapse
No announcement yet.

X-Plane 11 Flight Simulator With Vulkan Performing Very Well On Linux - NVIDIA/AMD OpenGL vs. Vulkan Benchmarks

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

  • X-Plane 11 Flight Simulator With Vulkan Performing Very Well On Linux - NVIDIA/AMD OpenGL vs. Vulkan Benchmarks

    Phoronix: X-Plane 11 Flight Simulator With Vulkan Performing Very Well On Linux - NVIDIA/AMD OpenGL vs. Vulkan Benchmarks

    Last week the X-Plane 11.50 beta was released with its long awaited Vulkan renderer to complement its mature OpenGL rendering code. Since then we've been busy benchmarking with 23 different graphics cards of AMD Radeon and NVIDIA GeForce line-ups while running Ubuntu Linux and comparing the OpenGL vs. Vulkan rendering performance for this realistic flight simulator.

    http://www.phoronix.com/vr.php?view=29051

  • #2
    Simulators are, I guess the place where we should see the biggest improvements with graphics API's

    Comment


    • #3
      Seems like it is heavily ROP and or Geometry culling limited... note how 5600-5700XT all perform basically the same because they have the same # of ROPs even though everything else aboubt them varies widely.... it seems they still have things to learn about using Vulkan in tier engine.

      Also Radeon 5500 has 32ROPS... so if you take about 72fps/40(Navi10 ROPS) *32 (5500xt rops) you get aobut 57fps which is about what it is doing....

      Also was PCIe 4.0 enabled for the Navi cards? Another potential bottleneck is 5500xt is 8 lanes and 5600-5700xt are 16x. So command latency is probably a little higher to the 5500xt... but that should only be a percent or so?
      Last edited by cb88; 04-08-2020, 08:36 PM.

      Comment


      • #4
        X-Plane will only work under GNOME/Wayaland. In other news, RedHat now owns phoronix.com

        Comment


        • #5
          The improvements for AMD were quite a bit larger on Windows and OSX, due to the terrible state of OpenGL drivers on those platforms. Still a nice improvement here too.

          Comment


          • #6
            I can't understand why the GPU names in graphs are printed vertically in a tiny font. I just can't bring myself to zoom in all the time and try to decipher it. On a mobile phone, I no longer even open such an article, it's a complete no-go. Please fix the graphs. Ever since the new graphs were introduced, my Phoronix visits dropped in half.

            Comment


            • #7
              X-Plane is primarily CPU bound, always has been. The hope is that with the vulkan renderer now working, draw call preparation can be made multi-threaded in the future. At the moment you need to test on a CPU that can boost 1-2 cores to 5GHz to get max perf out of a 2080Ti for example, and even then nvidias gpu utilization shows something like 40%!

              Comment


              • #8
                Hi Michael,
                thank you very much for your incredible articles. But this tests do not mean anything to me. Why? Because report Average FPS is irrelevant for describe the experience of play. You need to report 1% lows and 0.1% lows as explained here: https://www.youtube.com/watch?v=uXepIWi4SgM
                Also Average FPS is useless anyway because the most important parameter is Minimum. The best games are game that have fps above monitor refresh rate and low latency with vsync enabled (like Worlfenstein series, I enjoyed because they have very low latency with vsync enable you don't have tear and also no stutter).
                Your recent tests on wayland vs xorg game play is irrelevant also because I played games in both environment and even at high fps the stutter make it unplayable in this environments.

                You need to start measure accurate frame time and then you'll discover very bad things that happends with Xorg and Wayland compositors still not optimized and laggy.
                Last edited by sp82; 04-09-2020, 04:19 AM.

                Comment


                • #9
                  Why does the Titan RTX perform worse than other Nvidia card?
                  I always thought, here are the low-cost AMD gpus starting in the list, but there is written "Titan RTX"?

                  PS: No, I don't care about Nvidia performance, but could the results have been shifted by one line?

                  Comment


                  • #10
                    Seeing how this game is very much CPU bound, the more interesting question is now how the different combinations of GPU and CPU perform. Instead of doing a barrage of GPU's I would find it more interesting to see less GPU's and an additional CPU or two.

                    sp82 do you know how to get the individual frame times out of X-Plane (or min/max values)? The default logfile output only indicates the average over the run, nothing more.

                    I am running some local tests here too on a Ryzen 7 1800X + Radeon R580 with latest Arch Linux and mesa-aco-git from the AUR, but so far the Vulkan backend crashes. Since I am using the non-steam version I need to adapt the PTS test profile to work in this context as well.

                    Edit: by simply running the benchmark with the additional argument --verbose you'll get for each frame a CPU, GPU and total time.
                    Last edited by Davidovitch; 04-09-2020, 07:50 AM. Reason: found how to get frame times

                    Comment

                    Working...
                    X