Andre Müller escreveu:
Evgeni Golov wrote:
On Tue, 19 Jun 2007 14:01:08 +0200 Andre Müller wrote:
Probably a problem with the PM settings in the
ACPI subsystem of the kernel -- there are C-states
governors now. Without loading a governor, I got
some similar or same problem.
Do you mean CONFIG_CPU_IDLE and CONFIG_CPU_IDLE_GOV_*?
Can't tell ATM (I'm @ work), but that sounds quite right.
In menuconfig, it comes after the P-States governors :-)
The modules are called ladder (implementing classic behaviour)
and menu (a new one). I can't tell their merits in practice
apart, yet.
The 2.6.22-rc5-hrt1 kernel with CONFIG_CPU_IDLE=y and both "ladder"
and "menu" is working fine with my ThinkPad R50e and ICH4 chip.
The earlier problems with C-states not showing up in powertop and
power consumption higher than the vanilla kernel are gone. s2ram
is working (read bellow).
I get 8.2 Watts with both vanilla 2.6.22-rc4 and 2.6.22-rc5-hrt1 (on X).
As 8.2 W is low enough, I don't think it will ever be improved and
I am happy with it (plus than 5 hours on battery).
But the wakeups counting is impressive on 2.6.22-rc5-hrt1 :-)
Executing X with my patched Window Maker (which was responsable for almost
4 wakeups before the patch) I get _only_ 3.0 wakeups. Wow!!!
This is in an ultra-idle X, doing nothing but looking at powertop's list.
To get this amazing 3.0 wakeups I did a few things compared to my standard
usage:
1) I loaded the userspace (= 0 wakeups) governor instead of
ondemand (= 2.0 wakeups)
2) I killed the mouse service gpm (0.5 wakeups)
3) Disconnected the wireless link.
I have observed an issue with s2ram, tough. After coming back from s2ram
the power consumption is 0.5 W higher. I have rebooted 3 times and this
behaviour keeps the same. Does this also happen to you, Andre?
Carlos R. Mafra
_______________________________________________
Power mailing list
[email protected]
http://www.bughost.org/mailman/listinfo/power
