You are not logged in.
Hi, after upgrading to the 3.15 kernel the CPU clock frequency is very high even when in idle mode. Previously (in 3.14) it went down to 800MHz, now the i7z utility gives the following output (while htop shows no heavy processor usage):
Real Current Frequency 3539.79 MHz [100.00 x 35.40] (Max of below)
Core [core-id] :Actual Freq (Mult.) C0% Halt(C1)% C3 % C6 % Temp
Core 1 [0]: 3539.79 (35.40x) 1.06 97.9 1 0 28
Core 2 [1]: 3499.31 (34.99x) 1 100 0 0 27
Core 3 [2]: 3497.60 (34.98x) 1 99.3 0 0 25
Core 4 [3]: 3513.86 (35.14x) 0 100 0 0 26
This is reproducible on two of my machines (one being a laptop which is now constantly in turbo mode unnecessarily draining its battery). Everything regarding frequency scaling is set to the defaults, intel_pstate is enabled.
echo 20 > /sys/devices/system/cpu/intel_pstate/max_perf_pct
immediately sets the clock frequency of all cores to 800MHz, so I assume pstate is working.
The CPU models are i7-4770K and i7-4510U.
Any help or further information on this issue?
Offline
I can confirm that, my Xeon E3-1230v3 (haswell) keeps running at turbo clock (100x37) in idle
Real Current Frequency 3694.77 MHz [99.94 x 36.97] (Max of below)
Core [core-id] :Actual Freq (Mult.) C0% Halt(C1)% C3 % C6 % Temp
Core 1 [0]: 3685.23 (36.87x) 5.19 75.8 11.7 6.72 37
Core 2 [1]: 3667.91 (36.70x) 1 97 1 1 38
Core 3 [2]: 3668.83 (36.71x) 1.36 96.4 1.08 1 39
Core 4 [3]: 3694.77 (36.97x) 2.33 89.2 6.87 1.33 39
It seems that something went wrong with haswell CPUs and intel_pstate
cpupower frequency-info:
analyzing CPU 0:
driver: intel_pstate
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 0.97 ms.
hardware limits: 800 MHz - 3.70 GHz
available cpufreq governors: performance, powersave
current policy: frequency should be within 800 MHz and 3.70 GHz.
The governor "powersave" may decide which speed to use
within this range.
boost state support:
Supported: yes
Active: yes
Last edited by marcinz (2014-07-06 14:27:07)
Offline
I can confirm this issue, also on a Xeon 1230v3.
But I get the same results when using the 3.14.x series LTS kernel, so 3.15 shouldn't be the culprit.
Even if it was a stretch, I tried to see if the problem persists when not using any desktop environment but even then the problem persists.
Last edited by Aldur (2014-07-07 12:18:29)
Offline
i7z doesn't support Haswell, so those C state numbers are incorrect, use i7z-git.
As for the frequency, read this: https://bbs.archlinux.org/viewtopic.php?id=169409. Basically, there is no problem, what you're seeing is the busy freq, but the CPU isn't busy most of the time.
Edit: Another fine tool to use is turbostat.
Last edited by Gusar (2014-07-07 12:50:00)
Offline
Makes sense, thank you.
Offline