proc speed managers

Snarkout · 2006-11-10 18:18:05

I have a t43, and I want to setup speed stepping (or whatever) - I've seen numerous options, and want to know what the currently accepted "best" is. I want something automatic, but don't know whether the daemon/module approach is better or if the kernel builtins are better in terms of perforance vs battery life gain. Are any of them more "up to date" than others? Are any considered only useful on older hardware? I'm not asking for help setting things up, I'm asking for advice/suggestions - even the thinkwiki page essentially says "here's a bunch of options" as far as I can tell.

lucke · 2006-11-10 18:52:12

With the current state of kernel cpufreq governors, I think it's best to go with them. Stick to ondemand (or conservative) if you want to have the best performance and still save some power (and when not on battery) or powersave if you're really in the need of as much battery life as possible (i.e. when you're lost in the jungle with your notebook).

[edit: minor text corrections]

phrakture · 2006-11-10 21:45:03

lucke wrote:

or powersave, if you're really in the need of as much battery life as possible (i.e. when you're lost in the jungle with your notebook)..

Perfect advice, actually - the difference between powersave and ondemand is really minutes... and does that really make a difference in the long run? 10 or 20 minutes more?

Snarkout · 2006-11-11 02:07:56

Awesome - thanks a lot! I'll see what shakes loose and report back.

patroclo7 · 2006-11-12 12:59:14

phrakture wrote:

lucke wrote:
or powersave, if you're really in the need of as much battery life as possible (i.e. when you're lost in the jungle with your notebook)..
Perfect advice, actually - the difference between powersave and ondemand is really minutes... and does that really make a difference in the long run? 10 or 20 minutes more?

Well, the difference is minutes only if your system is not often under heavy load (if it is never under heavy load, then the difference can be nothing). Otherwise, with high loads on cpu, with ondemand you use the maximum speed and with powersave the minimum speed: this can have an influence on the battery life!

Snarkout · 2006-11-12 16:42:18

OK, I used cpufreqd and cpufrequtils, and have loaded a few different governors. Anyone know of a good way to keep track of your current cpu state other than cpufreq-info? The x86info plugin for gkrellm just bounces between 0kHz and 3 GHz seemingly randomly.

caved · 2006-11-15 09:31:55

Hey guys,
What i usually do is make sure i loaded up powernow_k7 module for my AMD mobile 1400+ (old) laptop and start the powernowd deamon, it has proven to run much much better and with a set and forget deamon.
It seaminglessly scales from 500 Mhz min to 1200 max with very low overhead , you wont even notice it (well only when i look at Conky ehe)

This may be personal but i feel that it runs much better then the cpu_freq thing i with ondemand govenor, it scales only from 700 to 1100 which is quit unacceptable in my case offcause. In short:
Am happy with Powernowd

Caved

Snarkout · 2006-11-17 00:40:48

Cool, I'll give that a look.

slubman · 2006-11-17 08:33:28

Snarkout wrote:

OK, I used cpufreqd and cpufrequtils, and have loaded a few different governors. Anyone know of a good way to keep track of your current cpu state other than cpufreq-info? The x86info plugin for gkrellm just bounces between 0kHz and 3 GHz seemingly randomly.

I would like to know if it's useful to use cpufreqd AND cpufreq, as I understand the thing, cpufreq just set the frequency once, where cpufreqd is a daemon to adjust the cpu frequency. So I can't see how usefull it is to start cpufreq then cpufreqd. Am I wrong ?

stmok · 2006-12-20 19:34:02

Sorry to dig up this thread, but I'm using an IBM ThinkPad R40 with a Pentium-M 1.4Ghz. The kernel is kernel26beyond (see in EXTRA repository)

I wasn't successful with cpufreqd. (wasn't able to reliably set the profile to "on demand", and it seemed to be stuck at 1.4Ghz...When I manually tried to adjust it to 600Mhz, sometimes it worked, other times it didn't and remained at 1.4Ghz)

When I installed powernowd, "on demand" CPU speed adjusting worked with minimal screwing around!

You just need to add: powernowd in DAEMONS section AND speedstep_centrino in MODULES section of rc.conf

That's it!

Arch Linux

#1 2006-11-10 18:18:05

proc speed managers

#2 2006-11-10 18:52:12

Re: proc speed managers

#3 2006-11-10 21:45:03

Re: proc speed managers

#4 2006-11-11 02:07:56

Re: proc speed managers

#5 2006-11-12 12:59:14

Re: proc speed managers

#6 2006-11-12 16:42:18

Re: proc speed managers

#7 2006-11-15 09:31:55

Re: proc speed managers

#8 2006-11-17 00:40:48

Re: proc speed managers

#9 2006-11-17 08:33:28

Re: proc speed managers

#10 2006-12-20 19:34:02

Re: proc speed managers

Board footer