COBB@brandeis.BITNET (Wes Cobb) (05/19/89)
I managed to get down to the Computer Cache in Quincy, Mass today and run some benchmark programs on a Turbo16 demo system they had there. These are the results we got running identical code on 3 different systems: Benchmark 1-Meg-520st Mega-4 Mega-4+Turbo16 --------------- ----------- --------- --------------- Vmips .66 .66 1.15 Graf 8.84s 9.37s 7.33s Savage 6.85s 5.71s 3.75s Whetstone 86195 96663 140447 Vmips is a basically just a >64K version of the Sieve with 32 bit ints throughout. Results are normalized so that a vanilla Vax-11/780 does 1.00 vmips. Graf is basically just a bouncing lines demo. Savage and Whetstone are IEEE single precision versions of the standard benchmarks with the same names. Graf, Savage, and Whetstone were compiled with Absoft Fortran V2.2 using Jos Vermaseren's improved runtime library. Vmips was compiled with Laser C. Wes Cobb Physics Dept Brandeis University