Score:1

cpufreq scaling governors "ondemand" and "conservative" won't ignore nice load on AMD Ryzen CPUs

fr flag

My kernel 5.11.0-22-generic automatically loads the cpufreq scaling driver acpi-cpufreq and the scaling governor schedutil for my Ryzen 9 5900X. I enabled the setting ignore_nice_load after changing the scaling governor to either ondemand or conservative in order to fit my needs.

Accordingly after the changes everything looks correct; cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor returns ondemand or conservative for all CPUs and cat /sys/devices/system/cpu/cpufreq/ondemand/ignore_nice_load returns 1.

Still after the change frequency scaling works as expected but the nice load is not ignored, i.e. BOINC processes running with a niceness of 19 cause an increase of the CPU frequency. Can you think of any way of ignoring nice load with a Ryzen 9 CPU?

Score:0
fr flag

This issue is fixed by the new kernel version 5.13.0-21-generic which I received after upgrading to Kubuntu 21.10. When ignore_nice_load is set to 1 the CPU frequency is not scaling with BOINC running in the background.

Still the background load has a significant effect on the CPU temperature. At idle state the CPU temperature jumps from a bit less than 30 degrees Celsius before starting the BOINC daemon to a bit more than 50 degrees Celsius after starting the BOINC daemon.

Setting ignore_nice_load to 0 or 1 has a visible impact on CPU frequency but not really on the CPU temperature and hence power consumption. ignore_nice_load only improves my CPU temperature by 10 degrees Celsius, from a bit more than 60 degrees Celsius down to a bit more than 50 degrees Celsius.

My takeaway is that CPU power management improved dramatically over the years and my CPU never really sits idle but instead saves energy by sleeping whenever possible. This strips away the relevance of a tool like BOINC which is supposed to leverage idle CPU cycles.

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.