No C states on powertop = no power reduction ?

[Len Brown]

n

On Mon, 16 Jun 2008, Frédéric BOITEUX wrote:

> 	Thanks for your answer,
> 
> Le ven 13 jun 2008 11:04:20 CEST, Arjan van de Ven
> <arjan at linux.intel.com> a écrit :
> 
> > Adding C-states, while not impossible, isn't all that easy either ..... especially on older systems
> > (on newer ones it's relatively easy because hte interface got standardized)
> 
> Actually I tried to add some C-states to the DSDT as I seen on the web
> (http://www.bughost.org/pipermail/power/2007-August/000890.html) but
> didn't succeed : powertop did show the C-states, but wake-ups per
> second's rate became an insane value (about 100 000) and use of
> C-states was always the same (about 50% in C0 and 50% in C2),
> even when modifying some known parameters to lower wake-ups rate.
> 
> The main difference between experiments is that their system did have
> already C1&C2, but mine didn't have any before to try to modify DSDT.
> So I wonder if updating Processor part of DSDT could be enough, or if
> the BIOS should be aware of C-states management, which mine lacks for
> sure.

This is generally a bad plan.
Even if the processor supports them, one has to wonder why the motherboard 
in their BIOS.  They may be working around a motherboard, chipset, or 
author of the web page above notes that the changes save power, but 
that the system occasionall hangs.  Is that really what you want to do?

The reason you get a bunch of wakeups per second it that the C-states you 
have invented do not actually exist, and thus return immediately.

-Len