Announcement

Collapse
No announcement yet.

Intel Xeon Platinum 8380 P-State CPU Frequency Scaling Comparison

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

  • Intel Xeon Platinum 8380 P-State CPU Frequency Scaling Comparison

    Phoronix: Intel Xeon Platinum 8380 P-State CPU Frequency Scaling Comparison

    As part of the curiosity-driven benchmarks and areas of technical interest now that we've gotten some of our initial Intel Xeon Platinum 8380 "Ice Lake" benchmarks out of the way has been looking into the performance of Linux's P-State CPU frequency scaling driver on the 3rd Gen Xeon Scalable server. Benchmarked for the latesting testing was the power/efficiency out-of-the-box with P-State powersave as used by default with many Linux distributions against the P-State "performance" mode as well as putting P-State into passive mode to be able to via intel_cpufreq to try the Schedutil governor that relies on the kernel's scheduler utilization data for making frequency scaling decisions. Here is a number of power/performance governor benchmarks with the dual Xeon Platinum 8380 server in these varying kernel configurations.

    https://www.phoronix.com/vr.php?view=30236

  • #2
    Interesting is schedutil, it manages to get both lower power consumption, and higher score then powersave. Powersave is total meme with max 720W, when performance is 640W. Something either is wrong here, or powersave is really horrible.

    Comment


    • #3
      Originally posted by piotrj3 View Post
      Interesting is schedutil, it manages to get both lower power consumption, and higher score then powersave. Powersave is total meme with max 720W, when performance is 640W. Something either is wrong here, or powersave is really horrible.
      P-State is a lie. powersave never saves power, and performance doesn't do too much either. And it doesn't have any other settings.

      I always disable intel_pstate and use CPUFreq instead, which has more settings that actually do the job (e.g. powersave puts the CPU at lowest clock).

      Comment


      • #4
        Could this be some kind of Ice Lake issue? I don't remember powersave being this terrible. On AMD, schedutil performs pretty much the same as performance with the exception for video encoding and even there the difference isn't as massive as this. I suspect that there is something up with ICL...

        Comment


        • #5
          Thank you very much for this benchmark. It clearly shows Ubuntu's default powersave is far from optimal when comes to performance. However, even such 'crippled' set up beats Windows by a large margin and tries to compete with DragonflyBSD.
          Last edited by Volta; 08 June 2021, 05:20 AM.

          Comment


          • #6
            Originally posted by Volta View Post
            Thank you very much for this benchmark. It clearly shows Ubuntu's default powersave is far from optimal when comes to performance. However, even such 'crippled' set up beats Windows by a large margin and tries to compete with DragonflyBSD.
            True, where is that BSDtard now?

            Comment


            • #7
              Michael Any chance you could include the classic ondemand at some point as well?

              Comment

              Working...
              X