You are not logged in.
Hi,
I have a Intel C2D E6300 on a Gigabyte GA-965P-DS3 Mainboard. Of course I would like to take advantage of the speedstepping capabilities. The Arch beginners FAQ suggests cpufreq (good idea, since it's in the kernel ).
My problem is, I dunno my processors min_freq and max_freq and couldn't find it online (looking through various onlineshops, intel's product page and google searches). And then I don't know whether that is for both cores together or for each core at a time.
I hope you can help me out with that one.
Then I came across acpid power throttling and so on and I was wondering whether cpufreq is still needed since acpid offers both power throttling and stepping throtteling.
That whole idea raised the question in me, what daemons you guys use (in general for everything, whole 'grep DAEMONS= /etc/rc.conf')..
[small]Yeah, I know, it's not good to cramb 5 different thoughts into one thread but my brain is fryed from class and they're all somewhat connected... well I guess, just please forgive me on that one this time :>[/small]
Offline
I use powernowd for frequency scaling. Don't let the name fool you, it works for Speedstep on intel chips too.
I like it because it's simple & doesn't require much configuration. Just make sure the cpufreq-userspace module is loaded at boot (along with any CPU specific driver) and have powernowd start in your DAEMONS line. It's default config is to scale your CPU's frequency up and down under load, ignoring niced processes.
OH, and it will determine the frequencies your processor can run at automagically. If you'd like to see what they are, as root, run powernowd -d -v. It also works for multiple processors/cores.
Sorry, edited again. I should also mention that cpufreq-info will show the same information since you probably already have it installed as a part of cpufrequtils
Last edited by dmartins (2007-04-20 12:54:41)
Offline
Thank you for your reply, dmartins.
Thank you for the tip with cpufreq-info (how could I have missed that? :>)! So it tells me my max_freq of 1.86GHz (correct) but a min_freq of 1.6GHz. That doesn't sound very correct to me. Anyhow, I entered that in my /etc/rc.d/cpufreq and issued a cat /proc/cpuinfo and it tells me that now one core of my two core processor runs at full speed and the other one is clocked down to 1.6GHz. lol, that's so fricken crappy. I think I'll try powernow one of these days.
Though, if someone got cpufreq running adequately on his/her Core2Duo processor, please let me know! thx
Offline
The cpufreq daemon is just a little script that calls cpufreq-set to set your minimum & maximum frequencies & the cpu frequency governor to use (which is a part of the kernel). The way the script is written it only sets up one of your processors/cores! That's what you experienced
It's easy to fix. As root, open /etc/rc.d/cpufreq. Look for these lines:
if [ "$params" != "" ]; then
cpufreq-set $params
stat_done
Change the cpufreq-set line to:
cpufreq-set -c 0 $params
And add a line below it:
cpufreq-set -c 1 $params
Now it should all look like this:
if [ "$params" != "" ]; then
cpufreq-set -c 0 $params
cpufreq-set -c 1 $params
stat_done
Now, for /etc/conf.d/cpufreq. I'd suggest commenting out the lines which set your minimum & maximum frequencies unless you actually want to set these to something other than what your processor's limits are.
As for the line which sets the governor to use, i'd choose either conservative or ondemand. The difference between them is pretty minimal, the conservative governor sits at your lowest freqency and increases step by step to the max when under load. It then drops straight back to the lowest. ondemand does the opposite, it jumps straight to the highest frequency under load then decreases step by step when idle. powersave stays at the lowest freqency and performance stays at the highest. userspace allows you or another program to control the frequency (this is what powernowd uses as the governor).
Once it's set up, cpufreq will probably work just as well as any other utility, it just isn't as configurable as some of them.
Last edited by dmartins (2007-04-20 21:03:59)
Offline
Because i run powersave to suspend, i let powersaved daemon take care of everyhing related to that sort of thing. It starts dbus, hal and acpid automatically, so there's no need for those in DAEMONS array.
Everything is configured in /etc/powersave/
Powersave daemon works great with my e6300! (on both cores ) Now if you are a member of "power" group you can change governor with gnome cpu monitor applet or another GUI tool.
"Your beliefs can be like fences that surround you.
You must first see them or you will not even realize that you are not free, simply because you will not see beyond the fences.
They will represent the boundaries of your experience."
SETH / Jane Roberts
Offline