feigin@batcomputer.tn.cornell.edu (Adam Feigin) (02/18/89)
Below are the results of a molecular modeling benchmark run on several machines. I didn't run all the benchmarks, just some, so I can't say a whole lot about the configuration of the machines. Of course, your mileage may vary. Flames to /nev/dull. Molecular Dynamics Benchmarks January 1989 ============================== Times are given to run 1,500 time steps for 108 Lennard-Jones atoms. Single precision; all floating point w/integer loop counter overhead; no I/O. All times are CPU times, arranged in ascending order. The program was written in very vanilla FORTRAN. Machine Time Speed MIPS (min:sec) (8250=1.0) (8250=1.2) --------------------------------------------------------------------------- Cray-1 00:06.8 82.4 98.9 [fully vect] IBM 3090-600 00:16.0 35.1 42.1 (opt(3) vector 00:33.1 17.0 20.4 (opt(3) IBM 3090-200 00:17.2 32.7 39.2 (opt(3) vector 00:35.1 16.0 19.2 (opt(3) FPS 264 00:36.0 15.6 18.73 Apollo DN10000 00:47.0 11.96 14.3 [opt] (1 CPU) Stellar GS1000 00:49.2 11.4 13.7 f77 -O2 Convex C210 01:08.2 8.23 9.88 [not fully vectorized] DECstation 3100 01:10.4 7.98 9.58 f77 -O2 VAX 8700 02:07.1 4.42 5.31 [VMS] Convex C1 02:10.3 4.31 5.17 [not fully vectorized] CSPI MAP-6420 02:50.2 3.3 3.96 VAX 8530 02:59.9 3.12 3.75 [VMS] VAX 6210 03:41.1 2.54 3.05 [VMS,est.] Sun 4/260 03:41.2 2.54 3.05 f77 -O VAXstation 3200 03:47.7 2.47 2.96 [VMS] VAXstation 3100 03:49.8 2.44 2.93 [VMS] Sun 3/160 06:17.6 1.49 1.79 [with fpa] VAX 8250 09:21.9 1.00 1.2 [VMS] MicroVAX II 14:36.8 0.64 0.77 [VMS] Sun 3/60 21:10.4 0.44 0.53 [no fpa] Sun 3/50 28:03.3 0.33 0.40 [ditto] ---------------------------------------------------------------------------