[gmx-users] athlon/xeon/opteron

[Bert de Groot]

Hi all,

since I've been bothering you from time to time with hardware questions, I 
thought I'd share the benchmark numbers I've collected with the list. I'm 
sorry for providing only numbers for our custom test system, and not for the
standard gromacs benchmarks, but this system is more representative for
a typical system that we are running (~80.000 particles, protein/lipids/water
using PME and for the rest mostly default mdp options).

numbers are for 1000 md steps (2 ps)
2s/4s means the shuffle/sort option of grompp was used
gromacs version: 3.1.4
LAM : 7.02 / usysv/ tcp_short=524288
fftw: 2.1.3


dual Athlon (2000 MHz/256Kb L2):
#CPU    time  speedup  scaling
1:      1092
2:       725   1.51    0.75
2s:      592   1.84    0.92


dual xeon P4 3.06 GHz/512kB L2, serverworks m/b chipset, 
SCI 2D low latency networking):
#CPU     time speedup scaling
1:        823
2:        504   1.63   0.82
2s:       418   1.97   0.98
4:        377   2.18   0.55
8:        270   3.05   0.38
16:       240   3.42   0.21


quad xeon P4 2.8 GHz (512K L2/2048K L3), serverworks m/b chipset:
#CPU      time speedup scaling
1:        872  
2s:       544   1.60   0.80
4s:       343   2.54   0.64


dual opteron 244 (1800 MHz/1024 KB L2): 
(thanks to Alex Ninaber from clustervison for a test account)
#CPU      time speedup scaling
1:        861 
2:        513   1.68   0.84
2s:       481   1.79   0.90


make your own decision, we made ours :-)

--

Bert

____________________________________________________________________________
Dr. Bert de Groot

Max Planck Institute for Biophysical Chemistry
Computational biomolecular dynamics group
Am Fassberg 11 
37077 Goettingen, Germany

tel: +49-551-2011306, fax: +49-551-2011089

email: bgroot at gwdg.de
http://www.mpibpc.gwdg.de/abteilungen/071/bgroot
____________________________________________________________________________