El jue, 10-05-2007 a las 18:17 +0200, Petter Reinholdtsen escribió: > [José L. Redrejo Rodríguez] > > I would also install powernowd or cpudyn (I prefer powernowd). It > > saves a lot of battery power, giving more time when unplugged . > > Why do you prefer such daemon instead of leaving that to the kernel to > handle? I notice the debian laptop task install cpufrequtils, and it > will feed values into /sys/devices/system/cpu/cpu0/cpufreq/* to enable > power saving mode when there is little to do. > cpufrequtils doesn't manage the cpu speed, it includes the tools to do it, but it doesn't do it. I guess you are refering to cpufreqd. Powernowd works smoother and faster than cpufreqd or cpudyn. I've tested, when starting some cpu consuming apps, how cpufreqd doesn't answer to this requirement as fast as powernowd does. And, in the opposite side, when no cpu power is needed powenowd slows down the cpu faster. So, the user experience is better and the battery safe is greater. Also, powernowd works with smp without needing any extra configuration, as cpufreqd needs. > > The problem with it is that you have to load different modules > > depending on your cpu to make it useful. cpufreq-userspace should > > always be loaded (for powernowd) and another module depending on the > > cpu: speedstep-centrino for most laptops without an amd core (a look > > to /lib/modules/`uname -r`/kernel/arch/i386/kernel/cpu/cpufreq would > > give hints). Nowadays, more than 90% of the laptops I see work with > > speedstep-centrino > > This problem is still present. Yes, you're right, > Are there any packages able to detect > and install the correct cpufreq kernel module present in Etch or > available from somewhere else? Not, I don't know of such package, but, loading by default speedstep-centrino doesn't hurt in an AMD laptop (well, it just says that's device doesn't exist). Doing it you cover most existing laptops. Also, a script that in the first start (after installation) of the system reads /proc/cpuinfo or just modprobe different modules finding out which one does not answer a error could easily be made. Regards. José L.
Attachment:
signature.asc
Description: Esta parte del mensaje =?ISO-8859-1?Q?est=E1?= firmada digitalmente