[mod.vlsi] SPICE Benchmarks from Prof Newton at UC Berkeley

greenberg@ELMER.DEC (Steve - HLO2-2/H13 Pole G13 225-6105) (12/04/85)

These results were compiled by Richard Newton at UC Berkeley; the example
circuit chosen was selected to expose a number of specific features of the
machines, like floating-point precision differences, and is one of the standard
SPICE benchmark circuits.

If you can add data for a commercially available machine that is not in this
list, I am sure Professor Newton would be pleased to add the data to this
table.

SPICE version 2G6 was the version that was used.  Time is normalized to a
VAX11/780 with Floating Point Accelerator, running UNIX, with optimizer turned
off in the FORTRAN compiler. TRAN TIME is the time for the transient anlysis
portion, TRAN ITER is the number of iterations in transient analysis,
/DEVICE/NR is the time per device per Newton-Raphson iteration.


                       WORKSTATION SPICE-2 COMPARISON
		27 MOSFET, 101 print points, MOSAMP2 Circuit

MACHINE        TOTAL    TOTAL    TRAN     TRAN   /DEVICE      COMMENT
               TIME     NORM     TIME     ITER     /NR
              (mi:se)            (sec)            (ms)

CRAY1/mc     0:02.42   0.014      2.16    426     0.187   m/c code gen.
CRAY1        0:02.78   0.016      2.52    449     0.207   no m/c code gen.
IBM3081K/mc  0:05.74   0.033      5.31    424     0.463   m/c code gen.
CYBER760/mc  0:05.85   0.035      5.24    385     0.504   m/c code gen.
CYBER760     0:06.44   0.038      5.85    385     0.562   no m/c code gen.
VAX8600      0:18.67    0.11     17.34    385     1.668   VMS4.1, FPA, FOR4.0
VAX8600      0:27.68    0.16     25.77    385      2.47   UNIX 4.3,FPA,new f77#
VAX8600      0:42.77    0.25     40.30    424      3.52   UNIX 4.3, FPA, not-O
VAX11-780       1:08    0.39     62.70    385      6.03   VMS4.1, FPA, FOR4.0
uVAXII          1:20    0.47     73.76    385      7.09   uVMS, FPU,FOR4.0
VAX11-785       1:21    0.48     75.08    385      7.22   UNIX, FPA, new f77#
SUN3/68881      1:35    0.56     89.70    385      8.62   16MHz 68020,UNIX
VAX11-785       1:41    0.60     94.35    424      8.24   UNIX, FPA, not-O
HP9000/550      1:51    0.66     99.62    385      9.58   HPUX?, FPA
APOLLO DN660-E  1:55    0.68    107.00    385      10.2   in-line FP code
VAX11-780       2:29    0.88    140.20    424      12.2   UNIX, FPA, -O (Utah)
uVAX II         2:45    0.98    154.20    424      13.5   ULTRIX, FPA, -O
uVAX II         2:49    0.99    157.75    424      13.8   ULTRIX, FPA, not-O

VAX11-780       2:49     1.0    159.12    424      13.9   UNIX, FPA, not-O

HP9000/orig     2:69     1.1    172.05    385      16.6   HPUX, FPA
MASSCOMP/FPA    3:29     1.2    192.6     385      18.5   UNIX, FPA, -O, IEEE
VAX11-750       3:33     1.3    200.20    424      17.5   UNIX, FPA, -O
APOLLO DN320    4:27     1.6    251.00    424      21.9   8Mhz 68010, bslice FPA
uVAX1/QVSS	7:25	 2.6	422.55	  385	   40.6   uVMS4.0
IBMPC/AT       10:57     3.9    606.37    445      50.5   DOS, 8287, some SP
SUN2-SKY       12:58     4.6    745.87    424      65.2   UNIX, FPA, -O

IBMPC/XT       16:07     5.8    904.18    465      72.0   DOS, 8087, some SP

MASSCOMP       16:15     5.8     935.6    415      83.4   UNIX, no FPA, -O
APOLLO DN300   17:51     6.3   1029.00    385      99.0   8MHz 68010, no FPA
HP9836         18:47     6.7   1081.18    385     104.0   HPUX, no FPA, IEEE
HP217FPA       22:45    8.07   1303.07    385    125.35   FP card
HP217          29:05    10.3   1675.93    385    161.22   no FPA
SUN2-IEEE      29:39    10.5   1713.67    385     164.8   UNIX, no FPA, -O, IEEE

Copyright (C) A.R.Newton, October 1985


Note: Although the uVAX II hardware is slower than the VAX11-780, the uVAX 
ULTRIX F77 compiler is 6% faster than UNIX.  The net result is that the uVAX II
running ULTRIX runs the same speed as the VAX11-780 running UNIX.

The following is the SPICE input file that was used in the above benchmarks:

mosamp2 - mos amplifier - transient
.WIDTH OUT=80
.OPTIONS ACCT ABSTOL=10N  VNTOL=10N
.TRAN 0.1US 10US
M1  15 15  1 32 M W=88.9U  L=25.4U
M2   1  1  2 32 M W=12.7U  L=266.7U
M3   2  2 30 32 M W=88.9U  L=25.4U
M4  15  5  4 32 M W=12.7U  L=106.7U
M5   4  4 30 32 M W=88.9U  L=12.7U
M6  15 15  5 32 M W=44.5U  L=25.4U
M7   5 20  8 32 M W=482.6U L=12.7U
M8   8  2 30 32 M W=88.9U  L=25.4U
M9  15 15  6 32 M W=44.5U  L=25.4U
M10  6 21  8 32 M W=482.6U L=12.7U
M11 15  6  7 32 M W=12.7U  L=106.7U
M12  7  4 30 32 M W=88.9U  L=12.7U
M13 15 10  9 32 M W=139.7U L=12.7U
M14  9 11 30 32 M W=139.7U L=12.7U
M15 15 15 12 32 M W=12.7U  L=207.8U
M16 12 12 11 32 M W=54.1U  L=12.7U
M17 11 11 30 32 M W=54.1U  L=12.7U
M18 15 15 10 32 M W=12.7U  L=45.2U
M19 10 12 13 32 M W=270.5U L=12.7U
M20 13  7 30 32 M W=270.5U L=12.7U
M21 15 10 14 32 M W=254U   L=12.7U
M22 14 11 30 32 M W=241.3U L=12.7U
M23 15 20 16 32 M W=19U    L=38.1U
M24 16 14 30 32 M W=406.4U L=12.7U
M25 15 15 20 32 M W=38.1U  L=42.7U
M26 20 16 30 32 M W=381U   L=25.4U
M27 20 15 66 32 M W=22.9U  L=7.6U
CC 7 9 40PF
CL 66 0 70PF
VIN 21 0 PULSE(0 5 1NS 1NS 1NS 5US 10US)
VCCP 15 0 DC +15
VDDN 30 0 DC -15
VB 32 0 DC -20
.MODEL M NMOS(NSUB=2.2E15 UO=575 UCRIT=49K UEXP=0.1 TOX=0.11U XJ=2.95U
+   LEVEL=2 CGSO=1.5N CGDO=1.5N CBD=4.5F CBS=4.5F LD=2.4485U NSS=3.2E10
+   KP=2E-5 PHI=0.6 )
.PRINT TRAN V(20) V(66)
.PLOT  TRAN V(20) V(66)
.END