Announcement

Collapse
No announcement yet.

Intel Coffee Lake OpenGL Performance On Windows 10 vs. Linux

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

  • Intel Coffee Lake OpenGL Performance On Windows 10 vs. Linux

    Phoronix: Intel Coffee Lake OpenGL Performance On Windows 10 vs. Linux

    For those curious about the state of Intel's open-source Mesa OpenGL driver relative to the company's closed-source Windows OpenGL driver, here are some fresh benchmark results when making use of an Intel Core i7 8700K "Coffee Lake" processor with UHD Graphics 630 and testing from Windows 10 Pro x64 against Ubuntu 16.04.3 LTS, Ubuntu with the Linux 4.16 Git kernel and Mesa 18.1-dev, and then Intel's own Clear Linux distribution.

    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
    ...Linux game ports? Yeah, ports are by definition slower since the porting process adds overhead. That's why you should benchmark native games instead when trying to compare Linux vs. Windows driver performance.

    Comment


    • #3
      This is great to see as originally Intel Gen 9 graphics on Linux were still noticeably slower than the Windows drivers at the time.
      Crappy launches- but fine wine, seems to be common among Intel/AMD

      Comment


      • #4
        Windows is still like 90% of the market so they get the most of the resources and most of the OEM interest.

        Comment


        • #5
          Originally posted by agd5f View Post
          Windows is still like 90% of the market so they get the most of the resources and most of the OEM interest.
          OK i will patch that... crappy launches - but fine development, seems to be Linux common
          Last edited by dungeon; 22 February 2018, 12:39 AM.

          Comment


          • #6
            Originally posted by Swiftpaw View Post
            ...Linux game ports? Yeah, ports are by definition slower since the porting process adds overhead. That's why you should benchmark native games instead when trying to compare Linux vs. Windows driver performance.
            While in practice, many "ports" are just a Wine layer, a true port, such as to SDL, doesn't have those issues as long as the devs take the time to benchmark, test and fix bugs to a reasonable extent.

            Comment


            • #7
              Originally posted by dragorth View Post

              While in practice, many "ports" are just a Wine layer, a true port, such as to SDL, doesn't have those issues as long as the devs take the time to benchmark, test and fix bugs to a reasonable extent.
              But even that SDL behave differently on Linux or Windows... that is still a layer, not like wine but is still a layer. So proper port would be to avoid these layers, that would have best performance and would be more responsive (i particulary hate SDL input lag )... but well, it is too much work and everybody likes to avoid too much work...

              Repeat after me - Simple DirectMedia Layer, Simple DirectMedia Layer, Simple DirectMedia Layer It is not so simple, but of course it is an layer
              Last edited by dungeon; 22 February 2018, 01:02 AM.

              Comment


              • #8
                Thanks Michael. I love these Windows Vs. GNU/Linux benchmarks. I'm looking forward to the upcoming Radeon benchmarks.

                Comment


                • #9
                  Originally posted by dungeon View Post

                  But even that SDL behave differently on Linux or Windows... that is still a layer, not like wine but is still a layer. So proper port would be to avoid these layers, that would have best performance and would be more responsive (i particulary hate SDL input lag )... but well, it is too much work and everybody likes to avoid too much work...

                  Repeat after me - Simple DirectMedia Layer, Simple DirectMedia Layer, Simple DirectMedia Layer It is not so simple, but of course it is an layer
                  If they want the best performance possible, they should port the games to Assembly. Nothing beats direct interaction with the CPU

                  Comment


                  • #10
                    Unigine's benchmarks were the most demanding OpenGL tests used and basically amounts to torture with the Coffee Lake UHD Graphics at 1080p... But for what it's worth, the Intel Mesa driver came in right behind the Windows performance, basically the same.
                    11% lower performance is not "the same" in my book.

                    Comment

                    Working...
                    X