Announcement

Collapse
No announcement yet.

Mesa 11.0.2 vs. Mesa 11.2-devel Benchmarks On Intel Skylake

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

  • Mesa 11.0.2 vs. Mesa 11.2-devel Benchmarks On Intel Skylake

    Phoronix: Mesa 11.0.2 vs. Mesa 11.2-devel Benchmarks On Intel Skylake

    For those currently running Ubuntu 15.10 or other similar Linux distributions powered off Mesa 11.0, here are some performance benchmarks comparing that release to the about-to-be-branched Mesa 11.2...

    http://www.phoronix.com/scan.php?pag...-11.2-SKL-Mesa

  • #2
    Hmm I dunno, these differences look more like the realm of variable uncertainty to me.

    Comment


    • #3
      Originally posted by theriddick View Post
      Hmm I dunno, these differences look more like the realm of variable uncertainty to me.
      Agreed, it seems too small to be anything other than statistical noise.

      Comment


      • #4
        It should be noted that PTS dynamically increases the number of runs until the results are statistically significant. (If my memory serves me right). So these results are indeed not just noise. The TF2 test being an outlier; seems theres some noise in the 11.0.2 test as indicated by the variance "bar". All in all a small but nice improvement it seems.

        Comment


        • #5
          Yep, the minimum is 3 runs each. If they differ, even more.

          @Michael: could you extend this test with some AMD Apus plz?? Long time no see...

          Comment


          • #6
            Well I guess it looks like on this hardware there aren't any serious performance regressions.
            Is there more to see than that?

            Sometimes developers may have to overcome a local minima, changes which facilitate future gains and greater functionality in short term, involve a hit.
            Engineering is all about trade offs, was there any reason to expect better performance Skylake specific optimisations for example which may have drawbacks in some benchmarks?

            Comment


            • #7
              Originally posted by tomtomme View Post
              Yep, the minimum is 3 runs each. If they differ, even more.

              @Michael: could you extend this test with some AMD Apus plz?? Long time no see...
              I will have some new Kaveri tests in the coming week or two.
              Michael Larabel
              http://www.michaellarabel.com/

              Comment

              Working...
              X