Bug 206517

Summary:	scaling_cur_freq very different from cpuinfo_cur_freq on Acer Predator Helios 500 PH517-61-R0GX
Product:	Power Management	Reporter:	Bernhard Rosenkränzer (bero)
Component:	cpufreq	Assignee:	Zhang Rui (rui.zhang)
Status:	CLOSED INSUFFICIENT_DATA
Severity:	normal	CC:	rui.zhang, trenn
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	5.5.1	Subsystem:
Regression:	No	Bisected commit-id:

Description Bernhard Rosenkränzer 2020-02-13 15:27:52 UTC

On an Acer Predator Helios 500 PH517-61-R0GX, scaling_cur_freq values can vary greatly from cpuinfo_cur_freq:

$ sudo cat /sys/devices/system/cpu/*/cpufreq/cpuinfo_cur_freq
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
3200000
$ cat /sys/devices/system/cpu/*/cpufreq/scaling_cur_freq 
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955
548955

Unfortunately, the performance I'm seeing on this box while compiling stuff makes me think scaling_cur_freq is reporting correctly and cpuinfo_cur_freq is telling lies.

This might be related to bug 203035

Comment 1 Thomas Renninger 2020-02-14 08:02:47 UTC

Best use:
cpupower monitor

You can also measure workloads, e.g. when compiling something:

cpupower monitor make

Not sure what distro you are using and what the name of the package or whether cpupower is packaged there... The sources, even it's userspace is part of the kernel git repo:
tools/power/cpupower

Comment 2 Zhang Rui 2020-06-29 12:16:23 UTC

why I can not see this file /sys/devices/system/cpu/*/cpufreq/cpuinfo_cur_freq?

please run "turbostat -o ts.out sleep 10" while compiling, and see what frequency is reported there.

Comment 3 Zhang Rui 2020-11-19 03:36:51 UTC

Bug closed as there is no response from the bug reporter.
Please feel free to re-open it if the problem still exists in the latest upstream kernel.