Addendum: apparently hyperthreading is not on, I was misled by something on a linux forum that I used to test.
On Wed, Aug 1, 2012 at 9:10 AM, Laurence Marks <L-marks at northwestern.edu> wrote: > I am benchmarking some new nodes (Intel(R) Xeon(R) CPU E5-2660 0 @ > 2.20GHz) using the standard mpi Wien2k benchmark, and would be > interested in comments/comparisons -- the online mpi benchmark seems > to be a bit outdated. Hyperthreading is turned on (currently not under > my control), and the IB is not working right so these numbers are only > for a single node with two quadcores with openmpi 1.4 (no idea how it > was compiled) and composer_xe_2011_sp1.11.339 > > Runs on 1 node > ==> 16 cores (hyperthreading used) > Maximum WALL clock time: 93.7425620555878 > Maximum CPU time: 88.3000000000000 > > ==> 4 cores > Maximum WALL clock time: 221.049882888794 > Maximum CPU time: 210.040000000000 > > ==> 8 cores > Maximum WALL clock time: 132.835557937622 > Maximum CPU time: 125.560000000000 > > For comparison, on my older Intel(R) Xeon(R) CPU E5410 @ 2.33GHz > ==> 8 cores > Maximum WALL clock time: 304.228610000000 > Maximum CPU time: 296.800000000000 > > N.B., the speed increase with hyperthreading is mainly in HAMILT/HNS > and may be misleading or might be real. > > -- > Professor Laurence Marks > Department of Materials Science and Engineering > Northwestern University > www.numis.northwestern.edu 1-847-491-3996 > "Research is to see what everybody else has seen, and to think what > nobody else has thought" > Albert Szent-Gyorgi -- Professor Laurence Marks Department of Materials Science and Engineering Northwestern University www.numis.northwestern.edu 1-847-491-3996 "Research is to see what everybody else has seen, and to think what nobody else has thought" Albert Szent-Gyorgi