Announcement

Collapse
No announcement yet.

AMD Performance On Linux 5.11 Remains Mixed Due To Schedutil With Frequency Invariance

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

  • #11
    Originally posted by jeisom View Post
    I just tested this on my 3800x and linux 5.10 and the temps went up about 9C while idling when switched to the performance governor. It might have gone up more, But I switched back after it got to 43. cpupower reported ~4GHz on all cores on performance as well idling.
    Temperatures are generally very jumpy with Ryzens, especially on Windows and even with balanced power plan. The reported clocks don't account power/clock gating, it says 4GHz for me too with the <10W.

    Originally posted by polarathene View Post
    I could be wrong but if you're not using pstate as a governor, it still uses it internally for the HW? I think passive mode is the name, which allows alternative governors to influence somewhat if they like?
    intel_pstate=disable disables the module entirely.

    Comment


    • #12
      Originally posted by intelfx View Post
      But when there _is_ work to do, cpufreq starts to matter, and the performance governor will force the CPU to highest frequency and thus your CPU will use considerably more power compared to a more conservative cpufreq governor.
      I've never been able to confirm that, apart from the more aggressive performance crippling governors/power plans. It's likely just a matter of a few minutes of difference of battery life with most tasks.

      Comment


      • #13
        Originally posted by aufkrawall View Post
        It's likely just a matter of a few minutes of difference of battery life with most tasks.
        No, it's not.

        Comment


        • #14
          There is something I did not understand from the article. Why does the schedutil algorithm not work well on specific AMD CPUs?

          If we are talking about an algorithmic problem it should manifest on all processors models and brands, right?

          Comment


          • #15
            Originally posted by zoomblab View Post
            There is something I did not understand from the article. Why does the schedutil algorithm not work well on specific AMD CPUs?

            If we are talking about an algorithmic problem it should manifest on all processors models and brands, right?
            Intel and AMD CPUs tend to do things a bit differently. Zen 2 and newer desktop AMD CPUs use aggressive clock gating, which means they force the CPU to start skipping clock cycles at low loads. For Zen 2 Desktop CPUs, the lowest possible clock is something around the 2.x GHz range. But if you use something like Ryzen Master (on Windows) that communicates with the CPU's SMU, you can find the "effective clock" will often be in the 200-800 MHz Range. So the frequency invariance governor might be getting confused by the CPU apparently being able to do different amounts of work at the same clock speed, messing up its load/boost calculations.

            Comment


            • #16
              Originally posted by intelfx View Post
              No, it's not.
              Either provide data or stop spamming.

              Comment


              • #17
                I think I will just stay on 5.10 LTS kernel until at least 5.12 or possibly longer, will monitor the situation. There is nothing significant in 5.11 which would warrant it's use instead of 5.10, really, for me at least - using Ryzen 1600X on my main machine. No need to fall into regressions or other complications.

                Comment


                • #18
                  Originally posted by piorunz View Post
                  I think I will just stay on 5.10 LTS kernel until at least 5.12 or possibly longer, will monitor the situation. There is nothing significant in 5.11 which would warrant it's use instead of 5.10, really, for me at least - using Ryzen 1600X on my main machine. No need to fall into regressions or other complications.
                  your 1600X (Zen1) should be fine and unaffected by these patches.

                  Comment


                  • #19
                    Originally posted by aufkrawall View Post
                    Either provide data or stop spamming.
                    For mobile use, every watt counts. For example, the Asus Zenbook has a 67 watt-hour battery, and gets nearly 14 hours of battery life. That means for general web-browsing the system is consuming 4.8 watts of power. Just 1 additional watt of idle power consumption would chop almost 2 hours of battery life away.

                    Comment


                    • #20
                      Originally posted by boxie View Post

                      your 1600X (Zen1) should be fine and unaffected by these patches.
                      Thanks! That's what I thought but wasn't sure.

                      Comment

                      Working...
                      X