dbr@cybvax0.UUCP (Douglas Robinson) (05/15/85)
Now that DEC has announced the MicroVax-II I thought that the net would like to see some benchmarks... You will need a 132 column printer for this (or a dense printing mode). The source to the benchmarks is available on request (PURE computational benchmarks - meant to run CORE resident on LSI-11 running VENIX-11). Please note the machine configurations at the end - they are important for comparing floating point results (FPA or NO FPA makes one hell of a difference!). Benchmark comparing COMPUTATIONAL SPEEDS of included CPU's VAX 11/780 MV-II VAX 11/750 MV-I 11/73 PC-AT 11/23+ size: (Bytes) char 1 1 1 1 1 1 1 short 2 2 2 2 2 2 2 int 4 4 4 4 2 2 2 long 4 4 4 4 4 4 4 float 4 4 4 4 4 4 4 double 8 8 8 8 8 8 8 sort: (Real,User) char (15.4,15.1) (17.5,17.2) (33.8,32.7) (53.6,50.1) ( 51.0, 50.0) (*) (115.0,113.9) short (11.2,11.0) (14.0,13.0) (25.4,24.4) (40.5,38.8) ( 39.0, 37.9) (*) ( 91.0, 90.2) int (11.8,11.6) (13.4,13.0) (25.9,25.0) (41.1,39.3) ( 39.0, 37.8) (*) ( 91.0, 90.2) long (11.8,11.7) (13.3,13.0) (25.9,24.9) (41.0,39.2) ( 92.0, 90.9) (*) (193.0,191.5) float (13.3,13.2) (16.4,16.2) (29.1,28.0) (45.7,43.9) (111.0,110.5) (*) (206.0,204.7) double (18.8,18.5) (19.3,19.0) (33.0,31.8) (66.4,64.0) (105.0,104.5) (*) (221.0,218.6) op: (Real,User) short + ( 2.4, 2.4) (2.7,2.4) ( 5.1, 4.6) ( 9.3, 8.4) ( 5.0, 5.2) ( 4.0, 3.4) ( 11.0, 10.7) * ( 4.3, 4.2) (3.9,3.7) ( 7.3, 6.8) (12.8,11.9) ( 8.0, 7.6) ( 5.0, 4.1) ( 17.0, 16.8) / ( 5.1, 5.0) (4.8,4.6) ( 7.4, 6.9) (13.0,12.1) (10.0, 9.5) ( 5.0, 4.9) ( 24.0, 23.5) int + ( 1.6, 1.5) (1.9,1.7) ( 3.7, 3.3) ( 7.5, 6.8) ( 6.0, 5.2) ( 4.0, 3.3) ( 11.0, 10.7) * ( 3.3, 3.2) (3.2,3.0) ( 5.6, 5.1) (11.0,10.1) ( 9.0, 7.7) ( 5.0, 4.2) ( 17.0, 16.8) / ( 4.3, 4.2) (4.2,4.0) ( 5.6, 5.3) (11.1,10.5) (10.0, 9.4) ( 5.0, 4.9) ( 24.0, 23.5) long + ( 1.6, 1.5) (1.9,1.7) ( 3.8, 3.3) ( 7.7, 6.8) ( 7.0, 6.7) ( 6.0, 5.3) ( 16.0, 15.3) * ( 3.3, 3.2) (3.2,3.0) ( 5.5, 5.1) (11.1,10.2) (32.0,32.1) (15.0,14.1) ( 67.0, 65.4) / ( 4.3, 4.2) (4.2,4.0) ( 5.8, 5.3) (13.7,10.6) (38.0,37.9) (21.0,20.1) ( 79.0, 77.8) float + ( 6.1, 5.8) (6.1,5.8) (11.8,11.2) (29.9,28.5) (26.0,25.1) (75.0,73.8) ( 53.0, 52.9) * ( 8.6, 8.4) (6.3,6.1) (19.7,19.0) (37.7,36.1) (39.0,37.7) (77.0,75.9) ( 91.0, 90.6) / (16.1,16.0) (6.9,6.7) (30.1,29.2) (55.7,53.8) (37.0,37.1) (84.0,82.8) (104.0,103.5) double + ( 5.6, 5.6) (5.4,5.2) (11.9,10.5) (27.9,25.9) (25.0,25.0) (80.0,79.4) ( 50.0, 50.5)** * (14.3, 8.3) (5.8,5.6) (19.4,18.3) (35.2,33.6) (39.0,37.7) (82.0,80.7) ( 88.0, 88.1)** / (28.1,16.0) (6.4,6.1) (29.3,28.3) (53.3,51.5) (38.0,37.1) (89.0,88.3) (102.0,101.0) sys: (Real,User,System) (11.0,.7,5.7) (6.5,0.4,5.9) (11.1,1.4,9.5) (20.2,2.9,16.8) (24.0,3.5,20.6) (9.0,4.1,4.9) (43.0,24.7,17.4) [ Real >> User indicates that some other job was running, use User for valid comparison ] * Not included due to compiler limitation of 8Kb stack space per routine (sort benchmark's require 32Kb stack!) ** Note that User>Real !?!?! System Descriptions: VAX 11/780 1.0 or 2.0 Mb(?!) mem, (NO FPA), RP05 System V R2 MV-II 5.0Mb mem, FPA, 2-RD52's ULTRIX-32m (beta 1.1-0 VAX 11/750 2.0 Mb mem, (NO FPA), 404Mb Wini (Fuji Eagle) 4.2BSD MV-I 2.0 Mb mem, RD52 (30Mb Wini) ULTRIX-32m (beta 1.0 R2) 11/73 J11-A: 512Kb mem, 160Mb Wini (CDC) Venix-11 PC-AT Enhanced: 512Kb mem, 20Mb Wini, 80287 Math Coprocessor Venix-86 11/23+ 512Kb mem, KEF11-A floating point, 404Mb Wini Venix-11 -- Doug Robinson Jobs don't kill programmers... programmers kill jobs! 617/492-8810 ...!{mit-eddie, harvard, mirror}!cybvax0!dbr