[gmx-users] Blue Gene/P
h.loeffler at dl.ac.uk
Tue Apr 1 10:10:09 CEST 2008
Ok, here just a few numbers.
System: 52.000 atoms, coarse-grained, 500.000 steps
BlueGene/P: 16(procs) x 4(threads) = 64 tasks (or whatever they are
8x4 7.25 h
16x4 11 h
32x4 > 12 h (queue only allows <= 12 h)
16x2 10 h
64x2 > 12 h
Note: the procs x threads figures above are the numbers I asked for. It
may be possible that the scheduler distributes in a different manner. I
That's all I have for the moment.
On Mon, 2008-03-31 at 17:29 -0700, Peter Tieleman wrote:
> Thanks. I've seen the BG/L argument, but BG/P processors are faster. I
> have no idea how much though, and am looking for actual numbers or if
> those are not available a guess based on actual BG/L numbers and some
> argument about scaling to a BG/P.
> Mark Abraham wrote:
> > Peter Tieleman wrote:
> >> Hi,
> >> Has anyone run benchmarks on a BlueGene/P ?
> > The word from IBM in December was that since GROMACS 3.x would lack
> > both assembly inner-loops and threading on BlueGene/L, that it wasn't
> > worthwhile running GROMACS on onw. I figure that goes for BlueGene/P too.
> > Mark
> > _______________________________________________
> > gmx-users mailing list gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > Please search the archive at http://www.gromacs.org/search before
> > posting!
> > Please don't post (un)subscribe requests to the list. Use the www
> > interface or send it to gmx-users-request at gromacs.org.
> > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> gmx-users mailing list gmx-users at gromacs.org
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
More information about the gromacs.org_gmx-users