https://bugzilla.kernel.org/show_bug.cgi?id=84851 Bug ID: 84851 Summary: CPU scaled down to 600 MHz even under load Product: Power Management Version: 2.5 Kernel Version: Tested at the moment with 3.17-rc5 Hardware: All OS: Linux Tree: Mainline Status: NEW Severity: normal Priority: P1 Component: cpufreq Assignee: cpufreq@xxxxxxxxxxxxxxx Reporter: bgamari@xxxxxxxxx Regression: No My Dell Latitude E7440 seems to occassionally enter periods where the CPU clockrate of all cores (as shown by cpupower and powertop) is stuck in the few-hundred-MHz range even under load. Often this is around 600 MHz. At the moment I'm running 3.17-rc5 although I believe I have observed similar behavior even under older kernels. I'm afraid I don't have a reliable technique for reproducing the issue although it may be correlated with S3 suspend. Various If under load for long enough (several minutes) it seems it will eventually clock back up to 2.0GHz. However, if I then kill the load it will clock back down to 600MHz (as expected) and remain there until again subjected to prolonged load. I've attached a tarball with output from the various performance monitoring tools (cpupower and turbostat) for several sets of conditions, * while the machine is idle (v3.15-rc5-idle) * while the machine is under load (provided by stress -c4) and clocked at around 600 MHz (v3.15-rc5-loaded) * after the machine has been load for roughly 3.5 minutes and has finally clocked up to 2 GHz. Oddly it doesn't make it above 2.0GHz despite cpuinfo_max_freq being 2.9GHz. (v3.15-rc5-loaded2) Looking through sysfs I found the following a bit odd, $ for i in /sys/devices/system/cpu/cpu0/cpufreq/*; do echo $i $(sudo cat $i); done cpufreq/affected_cpus 0 cpufreq/cpuinfo_cur_freq 628906 cpufreq/cpuinfo_max_freq 2900000 cpufreq/cpuinfo_min_freq 800000 cpufreq/cpuinfo_transition_latency 4294967295 cpufreq/related_cpus 0 cpufreq/scaling_available_governors performance powersave cpufreq/scaling_driver intel_pstate cpufreq/scaling_governor powersave cpufreq/scaling_max_freq 2900000 cpufreq/scaling_min_freq 800000 cpufreq/scaling_setspeed <unsupported> Note the peculiar value of cpuinfo_transition_latency. Seems just a tad long. -- You are receiving this mail because: You are the assignee for the bug. -- To unsubscribe from this list: send the line "unsubscribe cpufreq" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
