Announcement

Collapse
No announcement yet.

Windows 10 vs. Linux 4.15 + Mesa 17.4-dev Radeon Gaming Performance

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

  • #11
    Originally posted by marek View Post
    Third possibility: Neither. DX->OpenGL and DX->Vulkan are translation layers that always add inefficiencies. Windows drivers also probably have profiles for those games, so the Windows driver might not be running in its default state.
    How about recording the game on Linux via apitrace and benchmark that recording on Windows? Not sure if that is possible since the game may have Linux specific API calls that may fail on Windows but with some hacking it could work.

    Comment


    • #12
      These are pointless tests considering Windows is a DX11/DX12 platform, not OpenGL as the basis for their gaming experience. When Vulkan becomes the goto backend then show the outcomes.

      Comment


      • #13
        Originally posted by phoronix View Post
        the RX 580 was slower on Windows
        Do you mean Linux?

        Comment


        • #14
          I wonder what kind of framerate is being seen on Linux in windowed mode versus fullscreen compared to Windows? On Windows at Ultra quality settings @ ~1024x768 I'm getting between 80 - 110 FPS with a GTX 1070 and an i7-6700HQ.

          Comment


          • #15
            The test was made with Kernel 4.15 which has a known performance regression. https://bugs.freedesktop.org/show_bug.cgi?id=103100
            Also sisched can make some good difference.
            For me it boosts Shadow of Mordor from 68 to 72 and Tomb Raider 87 to 91 fps.
            Dont know if the tests where run with Performance Governor but usual that yields, depending on the test 1-2 fps.

            I think that the gap is still very big but it would be interesting to see a performance comparison without the buggy Kernel and with all available tweaks. On the other side, mesa has really to get rid of the fps drops.. It doesnt matter if the game runs with avg 70 vs 100 fps but the low minimum fps is a problem. Whats the problem here? Shader compiler?

            Comment


            • #16
              Hmm Blender linux vs Blender Windows ?

              Comment


              • #17
                Speaking of Windows, I played around with a wined3d for Windows a bit (screenshot) - just to find out that the same game runs more than twice as fast on Linux+Mesa. Of course both are nowhere near native D3D11 performance.

                And yes, that wined3d build does have CSMT enabled.

                Comment


                • #18
                  Little by little it keeps getting better and better every bench. The shadow of Mordor on rx580 is impressive. It equals windows while being a port oO. I can't imagine if the port was native. The results on F1 2017 are also pretty excellent, and it's a very good start for the first 100% vulkan port from feral.

                  Comment


                  • #19
                    So if an RX580 is $399 and the cheapest Vega64 is $848 (Vega56 $699), the 580 is a good buy at the moment?

                    Comment


                    • #20
                      Originally posted by Sluthy View Post
                      So if an RX580 is $399 and the cheapest Vega64 is $848 (Vega56 $699), the 580 is a good buy at the moment?
                      For the time being not at all I own a RX 580 8GB and on ubuntu 17.10 with kernel 4.13 and MESA 17.13 and its like I am using a 3yo entry level card... frame-drops are crazy avg FPS is low and generally its a waste of money if you just use linux... stick to nvidia they have way better drivers than whatever else is available for AMD cards.

                      Comment

                      Working...
                      X