I noticed that I could get my idle power (as reported by ACPI) down to 5.5W idle with a stock 2.6.34 kernel (the test I ran below didn't get it quite that low), but I have only been able to get down to 6.8W or so using 2.6.35 (despite having increased the time spent in sleep states). I found that with stock 2.6.35 the kernel would only go into C2, so I switched to using intel_idle, which added C4 state, but power didn't change much.
I also tried disabling smt and that added about 1.5W to idle power (despite even further reducing the wakeups-per-second below 10)
Is there any way to further identify where the power is being spent, or to figure out what is casuing the increased power consumption? On my netbook, the difference between 5.5W and 6.8 is ~1.5 hours (of course my average usage will be somewhat higher, but even when playing a movie, the difference is 7.1W vs 7.9W (still almost an hour of difference)
Here are some powertop numbers (computer completely idle, same power management settings applied, no wireless, lowest brightness:
2.6.32 (with ubuntu patches applied) SMT enabled:
Cn Avg residency
C0 (cpu running) ( 0.3%)
polling 0.0ms ( 0.0%)
C1 halt 0.0ms ( 0.0%)
C2 1.2ms ( 0.0%)
C3 30.0ms (99.7%)
P-states (frequencies)
1.67 Ghz 0.0%
1333 Mhz 0.0%
1000 Mhz 100.0%
Disk accesses:
Wakeups-from-idle per second : 33.5 interval: 20.0s
Power usage (ACPI estimate): 6.4W (7.8 hours)
Top causes for wakeups:
25.5% ( 13.2) [acpi] <interrupt>
18.1% ( 9.3) [kernel scheduler] Load balancing tick
17.9% ( 9.2) [kernel core] hrtimer_start (tick_sched_timer)
10.1% ( 5.2) events/0
8.0% ( 4.2) Xorg
4.5% ( 2.4) [Rescheduling interrupts] <kernel IPI>
4.1% ( 2.1) [extra timer interrupt]
2.6.34 stock kernel (smt enabled)
Cn Avg residency
C0 (cpu running) ( 0.3%)
polling 0.0ms ( 0.0%)
C1 halt 0.0ms ( 0.0%)
C2 0.4ms ( 0.0%)
C3 37.0ms (99.7%)
P-states (frequencies)
1.67 Ghz 0.0%
1333 Mhz 0.0%
1000 Mhz 100.0%
Wakeups-from-idle per second : 27.9 interval: 20.0s
Power usage (ACPI estimate): 6.1W (8.4 hours)
Top causes for wakeups:
27.0% ( 11.1) [kernel scheduler] Load balancing tick
15.7% ( 6.5) [Rescheduling interrupts] <kernel IPI>
12.7% ( 5.2) [kernel core] hrtimer_start (tick_sched_timer)
10.7% ( 4.4) [acpi] <interrupt>
10.0% ( 4.1) Xorg
7.2% ( 3.0) [extra timer interrupt]
2.6.35 stock kernel (intel_idle):
C0 (cpu running) ( 0.1%)
polling 0.0ms ( 0.0%)
C1 mwait 0.0ms ( 0.0%)
C2 mwait 0.4ms ( 0.0%)
C4 mwait 55.1ms (99.8%)
P-states (frequencies)
1.67 Ghz 0.3%
1333 Mhz 0..0%
1000 Mhz 99.7%
Disk accesses:
Wakeups-from-idle per second : 19.1 interval: 20.0s
Power usage (ACPI estimate): 7.0W (8.1 hours)
Top causes for wakeups:
24.4% ( 4.8) [kernel scheduler] Load balancing tick
22.6% ( 4.5) [kernel core] hrtimer_start (tick_sched_timer)
20.4% ( 4.0) Xorg
9.8% ( 1.9) gnome-terminal
Here is the same 2.6.35 kernel after disabling cpu1 (via echo 0 > /sys/devices/system/cpu/cpu1/online ):
Cn Avg residency
C0 (cpu running) ( 0.3%)
polling 0.0ms ( 0.0%)
C1 mwait 0.0ms ( 0.0%)
C2 mwait 3.2ms ( 0.0%)
C4 mwait 103.8ms (99.7%)
P-states (frequencies)
1.67 Ghz 1.7%
1333 Mhz 0.0%
1000 Mhz 98.3%
Disk accesses:
Wakeups-from-idle per second : 9.8 interval: 20.0s
Power usage (ACPI estimate): 8.3W (6.8 hours)
Top causes for wakeups:
28.4% ( 4.5) [kernel core] hrtimer_start (tick_sched_timer)
25.9% ( 4.0) Xorg
turning of the virtual CPU1 drastically reduced the number of "load balancing ticks" and gave me substantial time in C4 (103ms vs 55ms) but power jumped substantially.
Anyhow, I have no idea how to proceed. I have managed to get the cpu into lower sleep-states for longer periods of time, but my power reported by ACPI has been negatively affected.
Thanks,
.Geoff
_______________________________________________ linux-pm mailing list linux-pm@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/linux-pm
