Re: measuring cpu frequency
On 2004-05-01, Micha Feigin <michf@post.tau.ac.il> wrote:
> On Fri, Apr 30, 2004 at 04:48:24PM +0000, Ivan Fernandez wrote:
>> OK, I've been trying to set up cpu frequency scaling in my Dell Inspiron
>> 5150 (3.06GHz Mobile Pentium 4, debian testing and 2.6.5 kernel). I
>> seem to have everything in place, cpu frequency scaling compiled into
>> the kernel, together with performance and powersave governors, userspace
>> governor as default, frequency tables helpers too, and the p4-clockmod
>> module as driver. All the relevant files show up under
>> /sys/devices/system/cpu/cpu0/cpufreq/ and the userspace daemon I use to
>> switch frequencies (powernowd) loads and apparently changes frequencies
>> alright.
>>
>> My problem is that I can notice no cooling at all, CPU temperature is
>> around 70C most of the time, just as the same as before enabling the
>> whole cpufreq thing. I don't know how to measure real frequency:
>> the numbers that show on /proc/cpuinfo match with
>> /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed, but x86info -mhz
>> (which reads from /dev/cpu/0/cpuid, I believe, shows a different clock
>> speed (higher, either around 1500 MHz or, most of the time, the full
>> 3.06 GHz). Which one should I trust? Is there any other way to measure
>> cpu speed accurately?
>>
>
> I don't know how accurate it is but you could try bogomips. It doesn't
> report cpu frequency but the bogomips value which is a measure of the
> cpu speed (I don't know what the values are). It should let you know
> the relative speed of the cpu.
bogomips suggests that the right reading is the one on
/sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed, but I still don't
notice any cooling. I cut it down even more with
/proc/acpi/processor/CPU0/throttling (with both powersave cpufreq and
acpi throttling at 62%, bogomips are at 180, as opposed to 2084 at full
throttle and with the performance cpufreq governor). I have the
impression that the fan fires up less frequently, but no hard data on
this.
Thanks anyway
--
Ivan Fernández
ivan.fernandez@vanderbilt.edu
Reply to: