AMD Schedutil vs. Performance Governor Benchmarks On Linux 5.11 Shows More Upside Potential

[phoronix]

Phoronix: AMD Schedutil vs. Performance Governor Benchmarks On Linux 5.11 Shows More Upside Potential

With a pending patch, the Linux 5.11 AMD Zen 2 / Zen 3 performance is looking very good as far as the out-of-the-box performance is concerned when using Schedutil as is becoming the increasingly default CPU frequency scaling governor on more distributions / default kernels. With the previously noted Linux 5.11 regression addressed from when the AMD CPU frequency invariance support was first introduced, the Schedutil performance from small Ryzen systems up through big EPYC hardware is looking quite good. But how much upside is left in relation to the optimal CPU frequency scaling performance with the "performance" governor? Here is a look at those benchmarks on Ryzen and EPYC for Schedutil vs. Performance on a patched Linux 5.11 kernel.

http://www.phoronix.com/vr.php?view=29886

[andrei_me]

Awesome article Michael and with PTS the devs can reproduce this tests to check their changes

[pkunk]

Isn't CPUFreq not relevant for Intel?

[sdack]

The article says "schedutil", but several charts say "ondemand". Is this a mix up in the schedulers or a confusion in the naming?

[Michael]

The article says "schedutil", but several charts say "ondemand". Is this a mix up in the schedulers or a confusion in the naming?

Whoops, thanks, just a typo. As the Schedutil run is the same data from the other day in the Schedutil patching article, when I went to rename "Linux 5.11 Patched" to "CPUFreq Schedutil", I accidentally typed "CPUFreq Ondemand" instead. But as denoted by the auto-generated footnotes on the system table, the testing was indeed with Schedutil. Just a simple typo on my part with up until recently mostly seeing ondemand used.

[loganj]

is there any meaning is using schedutil when both performance and schedutil have almost the same power consumption?

[andrei_me]

is there any meaning is using schedutil when both performance and schedutil have almost the same power consumption?

My gut feeling is better idle efficiency

[pieman]

isn't intel post to switch to schedutil from powersave too?

[s_j_newbury]

Schedutil (and cpufreq in general) is supposed to save power under typical usage, not while running benchmarks... just saying.

[arQon]

ISTM that every single article about schedutil for the last God-only-knows how many years now could be summarised as "Better than the last version, but still absolutely nowhere near suitable for actual use". Except for the ones that could be summarised as "Worse than the last version, because we introduced an obvious bug but didn't bother running any tests at all before pushing it".

Wake me up when it starts to compete with even OnDemand, let alone Performance... :/