xeon 2.8GHz SMP/NOT test results

Kian Mohageri kian.mohageri at gmail.com
Mon Oct 16 11:15:30 PDT 2006


On 10/16/06, Attilio Rao <attilio at freebsd.org> wrote:
>
> 2006/10/16, Ivan Voras <ivoras at fer.hr>:
> > Kian Mohageri wrote:
> >
> > > I've never used sysbench (I essentially picked it randomly) so if you
> know
> > > it to be a crappy benchmark tool for this sort of thing, do tell.  I'm
> also
> > > pretty new at testing performance in general, but I hope someone finds
> it
> > > useful anyway.
> >
> > Maybe you'll be interested in ports/benchmark/unixbench, especially the
> > context switch and shell scripts benchmarks?
> >
> > > http://www.zampanosbits.com/smp_tests/
> >
> > Interesting results, especially for such an early version of the
> > processor (wrt HTT) - I'd expect much lower gain from HTT. While you're
> > at it, maybe you could add more results to your benchmark, like change
> > the timecounter to TSC, use various gcc optimization flags, twiddle
> > machdep.cpu_idle_hlt, use SMP kernel with HTT disabled in BIOS?
>
> What about PREEMPTION/FULL_PREEMPTION?



I'll take a look at doing some more tests like these with some actual
planning.  These ones were pretty much me getting bored at a coffee shop and
deciding to use our development server for testing.  I was curious about
disabling HTT in the BIOS too, after having read about it (that was peoples
suggestion, since supposedly it can hurt performance.)  One person said if
you don't disable it in the BIOS, `top` will only show CPU <50%, but I
didn't run into that, so maybe I'm misunderstanding them or it has been
fixed.

I'll do another round of tests with one or two of the other tools (lmbench,
unixbench) and options that you guys recommended and then publish the
results, I just have to find some time.  The tests themselves are pretty
quick but organizing the results takes a while.

I just have to be careful about what options I mess with so I don't lock up
the machine and have to head over to the server room :)

-- 
Kian Mohageri


More information about the freebsd-smp mailing list