[ltp] The zen of power saving, still 21.4W on Thinpad Z61p :(

[Marc MERLIN]

On Fri, Dec 28, 2007 at 08:10:22PM +0100, Laurent Gilson wrote:
> Try fixing the frequency at 1GHz. It will hurt performance... but you will
> not notice it in office. The > 600 switchs from 1GHz to 2GHz are a real 
> killer and should be avoided.

I was looking at this again:

I do consistently get 600-ish Wakeups-from-idle per second.

> Wakeups-from-idle per second : 585.5    interval: 10.0s
> Power usage (ACPI estimate): 22.5W (3.0 hours)
> 
> Top causes for wakeups:
>   22.3% ( 94.4)       <interrupt> : extra timer interrupt 
>   16.5% ( 69.7)       <interrupt> : HDA Intel, uhci_hcd:usb3, ohci1394, ipw3945
>   11.5% ( 48.5)      S20powernowd : queue_delayed_work_on (delayed_work_timer_fn
>   11.3% ( 47.9)       <interrupt> : acpi 
>   11.1% ( 46.9)            hdapsd : do_nanosleep (hrtimer_wakeup) 
>    7.5% ( 31.6)     enlightenment : schedule_timeout (process_timeout) 
>    4.8% ( 20.3)       <interrupt> : PS/2 keyboard/mouse/touchpad 
>    4.2% ( 17.7)              Xorg : do_setitimer (it_real_fn) 
>    3.4% ( 14.2)           gkrellm : schedule_timeout (process_timeout)

Not sure how to get rid of 'extra timer interrupt'. Any ideas?

For the second interrupt line, I unloaded a bunch of drivers:
tg3 usb_storage libusual ehci_hcd uhci_hcd usbcore irtty_sir sir_dev nsc_ircc irda ohci1394 sdhci mmc_core tifm_7xx1 tifm_core ipw3945

For acpi, I did: killall -STOP procmeter3 gkrellm hald-addon-acpi

Then killall -STOP hdapsd

Now, I'm left with:
> Cn                Avg residency       P-states (frequencies)
> C0 (cpu running)        ( 7.4%)         2.00 Ghz     0.0%
> C1                0.0ms ( 0.0%)         1.67 Ghz     0.0%
> C2                7.3ms ( 3.8%)         1333 Mhz     0.0%
> C3                6.4ms (88.8%)         1000 Mhz   100.0%
> 
> 
> Wakeups-from-idle per second : 144.5    interval: 10.0s
> Power usage (ACPI estimate): 18.7W (3.4 hours) (long term: 22.6W,/2.9h)
> 
> Top causes for wakeups:
>   36.9% ( 39.3)      S20powernowd : queue_delayed_work_on (delayed_work_timer_fn) 
>   30.9% ( 32.9)     enlightenment : schedule_timeout (process_timeout) 
>   10.2% ( 10.8)       <interrupt> : HDA Intel 
>    4.7% (  5.0)               tpb : do_nanosleep (hrtimer_wakeup)
>    2.8% (  3.0)     <kernel core> : queue_delayed_work_on (delayed_work_timer_fn)
>    2.3% (  2.5)       <interrupt> : acpi
>    1.8% (  1.9)              Xorg : do_setitimer (it_real_fn)

but as soon as I touch the keyboard, I'm back to 20-22W.

So, it looks like even going down to 150-200 wakeups per second doesn't
make any measurable difference.
Am I missing something here?

Thanks
Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems & security ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/