Announcement

Collapse
No announcement yet.

Patched Linux 5.11 Continues Looking Great For AMD Ryzen/EPYC Performance

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

  • Patched Linux 5.11 Continues Looking Great For AMD Ryzen/EPYC Performance

    Phoronix: Patched Linux 5.11 Performance Continues Looking Great For AMD Ryzen/EPYC Performance

    While the initial AMD Linux 5.11 performance regression written about at the end of last year was of much concern given the performance hits to AMD Zen 2 / Zen 3 processors with the out-of-the-box "Schedutil" governor, with a pending patch the regression is not only addressed but in various workloads we continue seeing better performance than even compared to Linux 5.10. Here is the latest from several more days of extensive performance testing.

    http://www.phoronix.com/vr.php?view=29881

  • #2
    More power usage is usually a good thing. Unless the CPU(s) are spinning uselessly It means more work is getting done. It's also nice to see the lower idle power use too.

    Comment


    • #3
      That's some incredible idle power saver, I guess based on max freq the scheduler clumps more tasks together in the idle state to enter deeper c states for the other cores.

      Comment


      • #4
        Incredible release for AMD owners.

        Comment


        • #5
          Originally posted by Zan Lynx View Post
          More power usage is usually a good thing. Unless the CPU(s) are spinning uselessly It means more work is getting done. It's also nice to see the lower idle power use too.
          Sort of - you don't want the power consumption going up disproportionately higher than the added productivity. In a lot of cases, the best results is getting more performance while power consumption doesn't increase as much. That is rare in a software change, but it does happen.

          Comment


          • #6
            Honestly, the results look great.
            I only have one question, or maybe a small request: how do the performance governor numbers match up against the 5.11 kernel+patch?

            That's be interesting to see, even with just a few benchmarks.

            Comment


            • #7
              As a reference there should have been a "performance" side-by-side.
              Without any other reference it's kinda hard to tell if it's doing a great job or just a less shitty job.

              Comment


              • #8
                Originally posted by BlueSwordM View Post
                Honestly, the results look great.
                I only have one question, or maybe a small request: how do the performance governor numbers match up against the 5.11 kernel+patch?

                That's be interesting to see, even with just a few benchmarks.
                It's coming in a follow up article... When running 200+ tests in some cases, can easily take 24 hours a run even on EPYC.
                Michael Larabel
                http://www.michaellarabel.com/

                Comment


                • #9
                  Originally posted by milkylainen View Post
                  As a reference there should have been a "performance" side-by-side.
                  Without any other reference it's kinda hard to tell if it's doing a great job or just a less shitty job.
                  It's coming in a follow up article... When running 200+ tests in some cases, can easily take 24 hours a run even on EPYC.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #10
                    Heh. Same request back to back. Tnx Michael.

                    Comment

                    Working...
                    X