[sclug] Under(re)volting - How does MS do it?

[ed]

On Wed, 28 Feb 2007 22:52:45 +0100
Pieter Claassen <pieter at claassen.co.uk> wrote:

> Hi Guys, I just got a brand new lenovo X60s from the Uni (which I am
> now regretting because it works so badly with Linux).
> 
> Many things don't work (most notably switching between external VGA
> and the laptop screen, increasing screen brightness, disabling
> wireless and bluetooth etc.).
> 
> But, what is really worrying is that the battery barely lasts 2 hours
> on this thing, even with dual cores (Intel Dual Core 1.66GHz with
> 2400MB Cache). The cpufreq governor works fine for both cpu's but
> only has three steps (1,1.33 and 1.66 GHz).
> 
> So, the question is simply: How come Microsoft manages to get around 6
> hours out of this battery (urban legend) while the best I can do
> (after manually disabling wireless, unloading USB drivers, enabling
> laptop mode) is only 2 hours? Yes, I know P=V^2/R which I suspect
> that even a minor drop in board voltage will have significant power
> gains. However, everything I read warns me against patching the
> kernel for undervolting because of possible hardware damage
> (motherboard not chip).
> 
> So, how is Microsoft doing this? How can I do this?

idle cycles on the cpu i imagine.

make sure one has cpu idle enabled in APM.

if you have ubuntu i think that's already enabled, but make sure you
have apm loaded at startup.

CONFIG_PM=y
CONFIG_APM=m
CONFIG_APM_DO_ENABLE=y
CONFIG_APM_CPU_IDLE=y

there are things like cpu scaling so the cpu cycles at a lower speed
when the battery is running low, or in little use. i don't know so much
about that as i tend not to need it, but i guess if everyone used this
they'd save a bit on the electricity bill per year.

-- 
The dirt trail to the moon pop is breakdancing because of the power of
the dark side. Qwest is blasting through 30 million dollars