Sun Archive Benchmark Page


Performance Index for several Sun3/3x machines:

The values are based on a modified version of dhrystone and give the relative performance compared to a Sun 3/50. The programm can be downloaded from here (57805 bytes).

The values are based on a modified version of whetstone and give the relative performance compared to a Sun 3/50. The programm can be downloaded from here (71664 bytes).

Both programs are compiled with cc at -O3 and static to always get the same conditions. rfloat requires a FPU.

For testing disk i/o performance, you can download the program Bonnie ((c) Tim Bray, 1990), either in source (19178 bytes) or binary (59827 bytes).

Compile it with "cc Bonnie.c -O2 -n -s -o Bonnie".

The source is modified not to execute the character i/o benchmarks, because these are almost CPU bound (e.g. 99% cpu time) on the ol' Sun3's. If you want them back, add -DCHARIO as parameter to cc.

The size of the test file should be about twice the main memory to get good results. If you use less, you'll measure cache performance, if you use more, you'll wait much longer and get almost the same numbers :-)

Performance Index for several Sun3/3x disks and controllers:

              -------Sequential Output-------- ---Sequential Input-- --Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec %CPU
3/260 XYN 200   242 99.1   334 24.7   205 25.7   223 99.0   580 35.7  29.1 16.6
3/260 XYN 100   242 99.1   335 25.3   207 26.6   224 98.3   609 37.2  33.1 17.5
3/260 XYN  50   242 98.9   338 24.7   221 26.5   226 97.7   693 40.6  38.9 19.9
3/260 XYN  25   238 96.8   351 25.0   249 28.2   232 97.6   844 42.8  55.4 25.4
3/260 XYN  12   244 99.4   373 24.6   322 32.5   248 98.1  1765 62.9 142.1 49.9
3/260 XYN   6   244 99.3   403 25.3   392 36.4   257 99.4  3233 99.9 263.5 83.7
3/260 RFN 400   255 99.1  1203 61.2   300 32.6   231 99.4  1404 96.8  25.4 13.0
3/260 RFN 200   253 98.6  1207 60.8   306 33.3   230 99.3  1407 94.9  27.7 14.1
3/260 RFN 100   256 99.3  1208 61.0   321 34.4   233 98.7  1441 91.5  31.1 15.7
3/260 RFN  50   256 99.3  1213 60.4   353 37.4   235 98.1  1491 86.2  36.7 17.6
3/260 RFN  25   258 99.5  1219 63.7   318 31.0   236 97.5  1532 83.6  51.8 22.9
3/260 RFN  12   256 99.4  1236 60.6   703 64.6   247 98.9  2363 87.3 119.2 41.6
3/260 RFN   6   256 99.2  1259 60.7  1101 99.9   254 99.2  3134 99.9 242.4 72.5
3/260 XYE  50   239 97.7   687 52.3   207 25.8   224 97.6   802 41.4  39.0 19.5
3/260 NFS  50   184 81.9   373 37.1   136 23.7   142 79.0   198 13.9  23.7 23.0
3/260 SD0  25   223 92.6   241 10.8   135 10.6   216 92.4   343 13.9  26.0 10.0
3/260 SD2  50   188 82.3   369 37.2   140 25.4   140 79.3   182 12.0  12.6 19.0
3/260 TMP1 50   255 97.3   367 14.4   215 12.5   156 62.2   371 13.8   9.7  6.0
3/260 TMP2 50   258 98.4  1069 47.7   485 36.6   198 78.5   762 27.6  19.0 11.5

The Sun 3/260 comes with 16MB main memory and 68020 at 25MHz.
I've tested a Xylogics 451 SMD disk controller (marked with XY), a Sun 2
SCSI controller (marked SD) and a Ciprio Rimfire SMD controller (marked RF).
N is a NEC 2362 9" SMD disk (670MB formatted).
E is a Fujitsu M2351 Eagle 11" SMD disk (380MB formatted).
0 is a Micropolis 1325 5.25" MFM disk with ACB 4000 (70MB formatted).
2 is a Micropolis 1558 5.25" SCSI disk (330MB formatted).
NFS is via ethernet (patched to do asynchroneous writes).
TMP1 was measured on tmpfs (2*32MB on two SCSI disks)
TMP2 was measured on tmpfs (65MB on one Fujitsu Eagle)
Peter Koch