on 07/12/2007 01:17 PM Bernhard Auzinger wrote the following:
>>>> bogomips        : 3611.22
>>>> TLB size        : 1024 4K pages
>>>>
>>>> bogomips        : 1603.39
>>>> TLB size        : 1024 4K pages
>>>>
>>>> 2) Just curious: Why is there such a big difference between the 2 CPUs'
>>>> bogomips values reported?
>>>> --
>>>> [EMAIL PROTECTED] mailing list
>> --
>> Piotr
>
> There has to be no difference between the two cores on dual-core processors. 
> That's because the bogomips-benchmark is a "calculated" benchmark that 
> depends mostly on the clock frequency and since the two cores on dual-core 
> athlons/opterons use the same clock, it's not possible to have different 
> values on that type of dual-core processor. I've absolutely identical values 
> (4825.77 bogomips) for both cores of my Athlon64 X2 4600+. 
>
> If you have two processors (f.ex. two single-cores) on your mainboard, than 
> you will have two slightly different values because of the slightly different 
> clock of the two processors. In your case it seems that the second processor 
> has a decreased clock. I'm sure you have enabled cool'n'quiet.
>
> Rgds

Actually I had never changed the CPUs freq/governor states, and they
were set to 800MHz/userspace.
After I changed their state using gnome applet, and from then on,  the
two CPUs' bogomips values  where equal.
For governor set to performance here they are:
#cat /proc/cpuinfo |grep -E 'proc|MHz|bogomips'
processor       : 0
cpu MHz         : 1800.000
bogomips        : 8125.24
processor       : 1
cpu MHz         : 1800.000
bogomips        : 8125.24

However, if I boot into M$ windblows (dual boot system), then boot into
linux without touching the CPUs' freq/gov state the MHz/bogomips values
are changed back to:
processor       : 0
cpu MHz         : 800.000
bogomips        : 3611.94
processor       : 1
cpu MHz         : 800.000
bogomips        : 1603.39

The 800 MHz is because I have compiled with
CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y
CONFIG_CPU_FREQ_GOV_USERSPACE=y


-- 
[EMAIL PROTECTED] mailing list

Reply via email to