Re: bogomips discrepancy on Intel Core2 Quad CPU

From: Bauke Jan Douma
Date: Sat Sep 08 2007 - 19:05:33 EST


Bill Davidsen wrote on 03-09-07 16:47:
Bauke Jan Douma wrote:
$> uname -a
Linux skyscraper 2.6.22.5 #7 SMP PREEMPT Sun Sep 2 12:12:25 CEST 2007 i686 GNU/Linux

$> cat /proc/cpuinfo | grep bogomips
bogomips : 4813.46
bogomips : 4810.91
bogomips : 4810.91
bogomips : 10583.94

The latter seems way off base.
Prod me for more info.

I see this occasionally on a dual, speedstep (or similar) finds a way to throttle down the cores under light load. I suspect that if you load the system:

for n in 1 2 3 4; do
nice -19 bash -c 'while true; do a=$RANDOM; done' &
done

Then you should see your bogomips rise on all cores.

Sorry for now getting back on this sooner. Other matters forced it to
the backburner.

I don't have Speedstep enabled on this machine's kernel (as a matter
of fact, the entire cpufreq Kconfig subtree is unchecked). Needless to
say running your bash script doesn't change the values in any way.

I'd say the 4800-something bogomips value is the correct one (for a
2.4GHz CPU -- at least that's what I recall on single-CPU machines,
mostly bogomips being approx. twice the rated CPU speed), and the
10000-something is the one off base.

Btw., the 10000 can be something like 12000-odd after a machine reboot,
and readout on a different CPU. The rest are all ca. 4810. Odd indeed.

bjd

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/