Announcement

Collapse
No announcement yet.

GeForce GTX 750 Series: Nouveau vs. NVIDIA Linux Driver Performance

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

  • GeForce GTX 750 Series: Nouveau vs. NVIDIA Linux Driver Performance

    Phoronix: GeForce GTX 750 Series: Nouveau vs. NVIDIA Linux Driver Performance

    With the forthcoming Linux 4.1 kernel there is finally out-of-the-box acceleration for the NVIDIA GeForce GTX 750 series on the Nouveau driver. With the Nouveau DRM/KMS driver able to self-generate the needed firmware/microcode to enable acceleration, it's quite easy to get running. However, the performance leaves a lot to be desired.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Originally posted by phoronix View Post
    Phoronix: GeForce GTX 750 Series: Nouveau vs. NVIDIA Linux Driver Performance

    The frame-rates with these GTX 750 graphics cards really isn't providing a playable experience.

    At least the hardware acceleration is easily working for the GeForce GTX 750 series, but the experience leaves a lot to be desired if you're a Linux gamer or looking to do more than just basic Linux desktop work with the current Nouveau driver.

    http://www.phoronix.com/vr.php?view=21673
    I think 70fps on 1080p with effects on High doesn't really qualify as "not playable". Nouveau is shy of the blob driver performance for various reasons, but no need to overstate things.

    However what wasn't mentioned is that some games show occasional artifacts (including Xonotic), like missing textures. And the fact that core profile ends up not being supported due to a missing GLX_ARB_context_create_profile is also a little annoying.

    Someone (unrelated to nouveau) recently RE'd the sched bits for Maxwell though, which is likely to be at least part of solving the rendering artifacts: https://github.com/NervanaSystems/ma.../Control-Codes.

    Comment


    • #3
      Looking at the fps, it is mostly low. For the perfomance the card have it is unplayable..

      Comment


      • #4
        Originally posted by sbolokanov View Post
        For the perfomance the card have it is unplayable..
        That doesn't make any sense. Anything above 30fps is playable, it's irrelevant that the card can do 200fps when properly clocked.

        A general note: For these nouveau vs. blob comparisons, until there's reclocking in nouveau, it would be interesting to test the blob configured to the lowest performance level. So you'd have nouveau vs. blob at lowest level vs. blob at full reclocking.

        Comment


        • #5
          Originally posted by Gusar View Post
          A general note: For these nouveau vs. blob comparisons, until there's reclocking in nouveau, it would be interesting to test the blob configured to the lowest performance level. So you'd have nouveau vs. blob at lowest level vs. blob at full reclocking.
          This. A benchmark of actual per-clock performance would be nice, instead of a "benchmark" that tells us nothing but: reclocking is not done yet in nouveau, so the blob is faster. Honestly, to find that out you don't need benchmarks, just common sense. As it is now, this article is nothing but a waste of electricity and database storage.

          Comment


          • #6
            Originally posted by MoonMoon View Post
            This. A benchmark of actual per-clock performance would be nice, instead of a "benchmark" that tells us nothing but: reclocking is not done yet in nouveau, so the blob is faster. Honestly, to find that out you don't need benchmarks, just common sense. As it is now, this article is nothing but a waste of electricity and database storage.
            Not only this article, but the 100 before it too, sadly ...

            Comment


            • #7

              Comment


              • #8
                Originally posted by MoonMoon View Post
                This. A benchmark of actual per-clock performance would be nice, instead of a "benchmark" that tells us nothing but: reclocking is not done yet in nouveau, so the blob is faster. Honestly, to find that out you don't need benchmarks, just common sense. As it is now, this article is nothing but a waste of electricity and database storage.
                maybe not to you, i found it interesting

                PS per-clock performance would also be flawed

                Comment


                • #9
                  ?

                  Originally posted by MoonMoon View Post
                  This. A benchmark of actual per-clock performance would be nice, instead of a "benchmark" that tells us nothing but: reclocking is not done yet in nouveau, so the blob is faster. Honestly, to find that out you don't need benchmarks, just common sense. As it is now, this article is nothing but a waste of electricity and database storage.

                  for what? nobody will underclock their card in real world, without some nvidia help nouveau will never work well, i use nvidia cards but i say now "nvidia" help nouveau kernel driver

                  Comment


                  • #10
                    Originally posted by imirkin View Post
                    However what wasn't mentioned is that some games show occasional artifacts (including Xonotic), like missing textures.
                    Thank for that info, that is most valuable one .

                    Article should be called: How broken rendering affect performance Really Michael, you shouldn't run any benchmarking if rendering is broken or just choose what isn't .

                    Comment

                    Working...
                    X