Re: SpeedStep,MHz,BogoMIPS,timing..

[Ari Heitner]

On Wed, 24 Jan 2001, Robert Dale wrote:

> 
> I bought a laptop which uses an Intel CPU with SpeedStep.  Sometimes when
> I boot the kernel recognizes the full MHz, and other times it is much less.
> And of course the BogoMIPS reflect this we well.
> 
> I'm wondering if this has any sort of side-effect on the kernel?  Will
> it disrupt timing loops and such? (I have no idea what I'm talking about ;)

My tp 600X has speedstep as well. When running on ac, the CPU goes full speed
(500MHz). On battery, it goes 1/4 speed (133Mhz) to save juice (must work, i
can run for 7 hours or so with an extra battery in the ultrabay).

linux seems happy with the switch -- the timing loops are close to right
(they're supposed to be order-of-magnitude). obviously if you boot the thing
then plug it in, the exact MHz count in /proc/cpuinfo will be wrong. But it's
never seemed a big idea.

You can get the same effect on an old 386 or 486 which still has a
"Turbo" switch.

> 
> Since Linus works at Transmeta, the kings of power saving, I'm guessing this
> has no effect.
> 

Quite the opposite :) SpeedStep is Intel's rather pathetic answer to Crusoe's
continuously variable powersaving (it's like having a two-speed car in
comparison to a slick continously variable transmission like Honda has been
working on iirc).

I love my 600X, and it blazes on wall power. But it took me 30 minutes t
compile the 8k lines Python-Visual today in class running on battery (g++ 2.95
with the optimizer on ... painful) :)

ari