Re: CPU power management
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Florian Reitmeir wrote:
> Hi,
>
> On Don, 15 Feb 2007, Eddy Petri??or wrote:
>
>> Eddy Petri??or wrote:
>>> Florian Reitmeir wrote:
>> sysctl dev.cpu.0.freq_levels
>>
>> Results in a list of frequencies down to 31 MHz. I tried some script I
>> found on the freebsd lists[1] to test the speed of blowfish at various
>> speeds and observed that on my Core 2 Duo that:
>
> i do really doubt this are useful settings, simply because..
>
> - to configure a cpu to some voltage and frequency
> -> you have to set some bits
> - there are limited combinations
> - the lowest setting is 0 0 0 0
> - the datasheets clearly states what this lowest setting means
>
> more states are possible, but clearly _not_ under the lowest value the cpu is
> capable. If your kernel states something like 31 MHz, this is simply an
The undervoltage patch does exactly that, make the processor run at
a lower speed than the designed minimal.
> error caused by some formular which interpolates from some bit settings and
> voltage settings the current cpu speed, instead of reading the MSR registers.
Then how do you explain that the speed of calculation has decreased
for speeds under the minimal one? The FreeBSD kernel *is*doing*
something to slow it down (not sure how healthy is for the CPU, but
that is another story).
> http://en.wikipedia.org/wiki/CPU_multiplier
The multiplier, AFAIK, is different from frequency scaling:
http://en.wikipedia.org/wiki/Frequency_scaling
- --
Regards,
EddyP
=============================================
"Imagination is more important than knowledge" A.Einstein
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
iD8DBQFF1RMKY8Chqv3NRNoRAoC7AJ4jtjTCIXCSlHkUmPSSaFhIqZ2A3wCgk2xw
sY0u8Xqsqt/ovoRmYfZ5i9M=
=lSPI
-----END PGP SIGNATURE-----
Reply to: