Why C3 state isn't entered?

[Bruno Ducrot]

On Thu, Nov 18, 2004 at 09:44:10AM -0800, Nate Lawson wrote:
> Bruno Ducrot wrote:
> >On Wed, Nov 17, 2004 at 01:22:33PM -0800, Nate Lawson wrote:
> >
> >>If you have the USB driver (usb.ko) loaded or compiled into the kernel, 
> >>you can't use C3.  The way to disable it is to implement support for 
> >>idling ports in uhci, ehci, and ohci.  C3 doesn't make a huge difference 
> >>(2-5%?) compared to C2 although it does help.  I did some profiling of 
> >>this a while back and found that the top three power saving features are 
> >>dimming the display (by far the most), changing CPU frequency (similar 
> >>but definitely less), and C2/C3 (better than C1 but not nearly as much 
> >>savings as the first two).
> >>
> >
> >Spin down disks is maybe more important than CPU frequency scaling also.
> 
> At least on my laptop, that didn't make much difference.  But we have 
> old behavior for many things like the syncer that is not power-aware so 
> spinning down disks may not be as helpful.

Well, it was more a general note.  Under FreeBSD, I guess you can save
something like 6 or 7% of battery remaining times if a more power
aware developpement is done for DD.  I don't expect much more.  Under
Linux, there is a need to do much more tweaks to achive a 20-30%
remaining time (via the laptop-mode) due to agressive disk access in
general.

> BTW, interested in implementing S4-OS for FreeBSD?  :)

I can't promise anything yet.

Cheers,

-- 
Bruno Ducrot

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.