An Early Performance Regression Hitting Highly Threaded Workloads On Linux 6.14-rc1

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • phoronix
    Administrator
    • Jan 2007
    • 67370

    An Early Performance Regression Hitting Highly Threaded Workloads On Linux 6.14-rc1

    Phoronix: An Early Performance Regression Hitting Highly Threaded Workloads On Linux 6.14-rc1

    With Linux 6.14-rc1 released I have begun trying out the new development kernel on a few systems locally. At least for high core count hardware tested thus far, Linux 6.14 at the moment during this early testing phase is sporting some performance regressions within some multi-threaded workloads.

    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
  • Danny3
    Senior Member
    • Apr 2012
    • 2408

    #2
    This is the real performance regressions finding that save billions of dollars, not that russian crap they traditionally lie about!

    I wish the fucking rich Linux foundation would support Phoronix properly as none other does such an important job, which indeed saves lots of money and power consumption, which in turn saves the planet a bit too.
    Very good job Michael again!

    Comment

    • schmidtbag
      Senior Member
      • Dec 2010
      • 6618

      #3
      I agree with Danny3 - this isn't the first time Michael has caught performance regressions early on. In this case, some of those regressions were quite severe.

      Comment

      • sophisticles
        Senior Member
        • Dec 2015
        • 2594

        #4
        Originally posted by Danny3 View Post
        This is the real performance regressions finding that save billions of dollars, not that russian crap they traditionally lie about!

        I wish the fucking rich Linux foundation would support Phoronix properly as none other does such an important job, which indeed saves lots of money and power consumption, which in turn saves the planet a bit too.
        Very good job Michael again!
        What the hell, I'll take the bait:

        How exactly does finding a performance regression in the Linux kernel result in saving billions of dollars?

        Comment

        • sophisticles
          Senior Member
          • Dec 2015
          • 2594

          #5
          Originally posted by schmidtbag View Post
          I agree with Danny3 - this isn't the first time Michael has caught performance regressions early on. In this case, some of those regressions were quite severe.
          So you agree that finding the regression results in saving billions of dollars?

          And it also saves the planet?

          What is it saving the planet from?

          Comment

          • Phoronos
            Senior Member
            • Mar 2024
            • 170

            #6
            Thanks Michael !!

            Comment

            • Danny3
              Senior Member
              • Apr 2012
              • 2408

              #7
              Originally posted by sophisticles View Post

              What the hell, I'll take the bait:

              How exactly does finding a performance regression in the Linux kernel result in saving billions of dollars?
              I was a reference to this claim:
              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

              And to the fact that regressions make the software run slow, which means it will take more time, which means it will consume more power.
              And will also require more hardware devices / nodes that cost money to purchase and will also consume more power.

              Originally posted by sophisticles View Post

              So you agree that finding the regression results in saving billions of dollars?

              And it also saves the planet?

              What is it saving the planet from?
              From faster life extinction because of the pollution that is created to keep up with the super computers and data centers energy demands.
              We are still very far away from producing electricity in 100% clean manner.
              While nobody wants to turn off the computers, at least we should make sure that they are not consuming more than supposed to.

              Comment

              • archkde
                Senior Member
                • May 2019
                • 687

                #8
                Originally posted by Danny3 View Post
                From faster life extinction because of the pollution that is created to keep up with the super computers and data centers energy demands.
                We are still very far away from producing electricity in 100% clean manner.
                While nobody wants to turn off the computers, at least we should make sure that they are not consuming more than supposed to.
                Also from premature replacements or addition of new computers. A surprisingly high percentage of the pollution linked to computers (exact numbers depend on device category, usage and how you weigh different kinds of pollution, but usually more than half) is actually not caused by their electricity consumption, but their manufacturing and (to a lesser extent) disposal.

                Comment

                • devling
                  Junior Member
                  • Dec 2022
                  • 14

                  #9
                  Originally posted by sophisticles View Post

                  What the hell, I'll take the bait:

                  How exactly does finding a performance regression in the Linux kernel result in saving billions of dollars?
                  10% slower, for example, equals 10% more power usage to get the same result (ok, I know it's not linear, but something like that). Multiply this by every server used by for example AWS, Facebook, Microsoft etc and that is a crapload of electricity and unnecessary extra resources being used due to buggy software. While it's hard to estimate the exact cost, it for sure it a ton of money.

                  Comment

                  • CommunityMember
                    Senior Member
                    • Oct 2019
                    • 1378

                    #10
                    The real question is whether the "will-it-scale" tests, which are an existing part of the Linux kernel CI processes, show the regression? If so, this report is just a duplicate. And if not, then will-it-scale should receive a PR to add in additional tests.

                    Comment

                    Working...
                    X