How to measure power consumption on a non-laptop system
Alan Jenkins
alan-jenkins at tuffmail.co.uk
Mon Oct 22 11:41:05 PDT 2007
Peter Hüwe wrote:
> Hi,
>
> is there an easy way to figure out the power consumption of my machine, which
> is a normal desktop pc ?
>
No.
There are power meters you could plug your computer into, which is
probably about as easy as it gets.
CPU's with frequency scaling advertise claimed power usage in
/proc/acpi/processor/CPU0/performance.
I don't think there's an equivalent for any other component, though if
you could see if other people have measured power savings with
components which match yours. E.g. if you're considering the ATA Link
Power Management patches, someone may have measured and published power
savings for ATA hardware similar to your own.
If you have a good CPU temperature sensor and disable "smart fan
control" (BIOS option), you can use the CPUs temperature as an
approximate, relative, non-linear measure of its power usage during
experiments.
e.g. I use highly inadvisable kernel patches to undervolt my CPU,
lowering the voltage used at maximum frequency to the default used
during the minimum frequency. I'm not suggesting you try undervolting,
but I compared the temperatures with the CPU under load, and I did see a
difference. You should see something similar with enabling CPU
frequency scaling.
Alan
More information about the Power
mailing list