Re: Linux SMP performance vs. UP: not a laurel

Dr. Kelsey Hudson (kernel@blackhole.compendium-tech.com)
Thu, 20 Jul 2000 15:45:33 -0700 (PDT)

You do realize that unless you have the client(s) process the same
workunit, your tests are skewed anyways...Some workunits take much more
time to complete than others.

FWIW, it takes my SMP 2xPIII/600B 128MB RAM 7 hr 47 minutes on average to
complete 2 units.

It takes my Sun Ultra 10 (UltraSparc IIi 333MHz) 768MB RAM 7 hr 20
minutes on avg. to complete one unit.

I homestly can't see your problem; Could you elaborate please?

Kelsey Hudson khudson@ctica.com
Software Engineer
Compendium Technologies, Inc (619) 725-0771
---------------------------------------------------------------------------

On Tue, 18 Jul 2000, Mario Vanoni wrote:

> UP PII333 overclocked to 412MHz, 512MB memory:
> 24/7 one setiathome <8hr 45min constant
>
> UP PII450 512MB memory, production machine:
> 24/7 one setiathome <8hr 10min constant
>
> UP PIII550 512 MB memory, no load:
> 24/7 one setiathome <7hr 00min _VERY_ constant
>
> All UP's 2.2.16, glibc-2.1.3, gcc-2.95.2, base SuSE 6.4.
>
>
> SMP 2xPIII550 1GB memory, mostly unloaded:
> 24/7 _TWO_ setiathome running, time for the single seti:
>
> 2.2.7, libc5, gcc-2.7.2.3, base SuSE 6.1:
> <8hr 50min, inconstant with values <7hr 50min
> 2.2.14, glibc-2.1.1, gcc-2.95, base SuSE 6.3:
> >9hr 15min, inconstant with values >10hr 10min
> 2.2.16, glibc-2.1.3, gcc-2.95.2, base SuSE 6.4:
> same as for 2.2.14, with some stategies like
> memtest before/after setiathome starting,
> mmap002, connections via ethernet 10M/RG-58,
> telnet from other machines, find | cpio to tape
> and others not noted, the time goes down:
> <9hr 20min, but never constant as with the UP's.
> 2.4.0-test3 after reboot (sic!) >10hr 45min !!!
>
>
> Now the SMP runs 2.2.16, I need it compatible for work.
>
> Any _REAL_ and usable explication?
> ==================================
>
> Willing to use the SMP for tests with seti.
>
> Not in list, only reader & UNIX sysadm/usr since AT&T SVR2.
>
> Regards
> Mario
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.rutgers.edu
> Please read the FAQ at http://www.tux.org/lkml/
>

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