Announcement

Collapse
No announcement yet.

NVIDIA's Linux Driver Can Deliver Better OpenGL Performance Than Windows 8.1

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

  • NVIDIA's Linux Driver Can Deliver Better OpenGL Performance Than Windows 8.1

    Phoronix: NVIDIA's Linux Driver On Ubuntu 14.10 Can Deliver Better OpenGL Performance Than

    Yesterday we looked at the Windows vs. Ubuntu 14.10 Intel OpenGL performance using Haswell HD Graphics to kick off our latest round of Windows vs. Linux OpenGL driver benchmarking. Out today is now our NVIDIA GeForce graphics card comparison on Windows 8.1 and Ubuntu 14.10 using the latest NVIDIA binary drivers. NVIDIA's new GeForce GTX 970 and GTX 980 Maxwell graphics cards were tested as well as the mature GTX 780 Ti.

    http://www.phoronix.com/vr.php?view=21108

  • #2
    The only one that was really interesting was the Unigine Tropics 1.3 benchmark. I wonder why the Tropics benchmark was such the outlier.

    All the others were essentially tied; or synthetic benchmarks that are mostly pointless.

    Regardless, good to know the OpenGL performance on NVIDIA is as good or better than Windows. I'm looking forward to when AMD's new driver strategy pays dividends and we can have good comparisons between AMD and NVIDIA on Linux!

    Comment


    • #3
      Michael, your comment about the Intel HD benchmark results from your earlier article was a little strange. There was only place where the Windows 8 driver was "largely leading the way", excepting the synthetic benchmarks which are now pretty close, that being the low-quality Xonotic test, which was probably due to the fact that the drivers are very much unoptimized for producing as many frames as possible with simple scenes and hit "CPU limit" as the CPU isn't waiting for the GPU. It's something that's hopefully being improved as you recently wrote about.

      Comment


      • #4
        Originally posted by adler187 View Post
        The only one that was really interesting was the Unigine Tropics 1.3 benchmark. I wonder why the Tropics benchmark was such the outlier.
        Because the windows kernel has significant overhead compared to Linux, and this becomes evident as the fps increases. The drivers themselves are still a bit more mature on windows side, and therefore at low fps windows becomes competitive. Alternatively it is the newer driver version that was used on windows that causes the performance increase when not limited by kernel overhead.
        Last edited by varikonniemi; 10-31-2014, 02:54 PM.

        Comment


        • #5
          Originally posted by s_j_newbury View Post
          Michael, your comment about the Intel HD benchmark results from your earlier article was a little strange. There was only place where the Windows 8 driver was "largely leading the way", excepting the synthetic benchmarks which are now pretty close, that being the low-quality Xonotic test, which was probably due to the fact that the drivers are very much unoptimized for producing as many frames as possible with simple scenes and hit "CPU limit" as the CPU isn't waiting for the GPU. It's something that's hopefully being improved as you recently wrote about.
          They can optimize OpenArena too, those numbers looks pretty low for HD4600 on both Windows/Linux Probably there is some regression there according to last year for example this one looks like more where it should be:

          http://www.phoronix.com/scan.php?pag...00_linux&num=4

          Comment


          • #6
            So as of right now the closed source nvidia drivers and the open source intel drivers are on par or superior to windows, and we haven't even switched to wayland yet. AMD I'm sure will be behind, but will catch up eventually - they're making good progress and the new amdgpu driver should really help speed things up on their end.

            Comment


            • #7
              These tests are meaningless on unigine- he should run dx11 and opengl on windows.

              Comment


              • #8
                Originally posted by sireangelus View Post
                These tests are meaningless on unigine- he should run dx11 and opengl on windows.
                You couldn't be more wrong there. Why compare apples and bannanas when you can compare apples to apples?
                Is it likely that Unigine is more optimized on DX? Yes.

                Tests has proven though that OpenGL and DirectX are on par performance wise while both do pretty much the same thing albeit in slightly different manner.
                That is completely unrelated to this test though, which compares the very same codepaths on two different systems.

                In fact, most of NVIDIA's driver code is exactly same on both systems, with exception of kernel interfaces that work with hardware.

                Comment


                • #9
                  Originally posted by sireangelus View Post
                  These tests are meaningless on unigine- he should run dx11 and opengl on windows.

                  And why? This test wouldnt be fair at all. DirectX is M$ only so why would anyone care about those tests. no one will care, no one cares.

                  Comment


                  • #10
                    Originally posted by gotwig View Post
                    And why? This test wouldnt be fair at all. DirectX is M$ only so why would anyone care about those tests. no one will care, no one cares.
                    No one cares about these tests either, these are just useless, we need some real life usage, so opengl sucks, besides, there is no good games for linux, plus nvidia doesnt care about opengl for windows, there is much better thing - directx, plus nvidia broke opengl performance with latest driver series, so yeah, so much of these "tests". When there will be tests of shadows of mordor, then we will talk, now even vsync ir broken on linux, who cares about some fps in fake tests...

                    Comment

                    Working...
                    X