Memory Bandwidth STREAM Benchmark John D. McCalpin john@mccalpin.com Revised to: Wed Feb 8 14:23:59 1995 MFLOPS Machine SSCAL Sum SAXPY --------------- ------ ------ ------ Cray YMP/C90 16 cpu 6541.0 4239.0 8651.1 Cray YMP/C90 8 cpu 3462.0 2535.1 5269.1 Cray T3D 256 PEs assembly 5264.3 2400.9 4673.2 Cray T3D 256 PEs Fortran 5265.1 1992.7 3770.7 CM-200 64k 10 (est) 2625.7 1977.1 3768.8 Cray YMP/C90 4 cpu 1736.8 1443.1 2920.3 CM-2 64k 8 MHz (est) 1912.0 1396.7 2666.7 Cray T3D 128 PEs assembly 2632.1 1200.5 2336.0 CM-2 64k 7 MHz (est) 1612.2 1219.1 2329.9 Cray Y/MP 8 cpu 1205.9 1107.9 2233.5 Cray T3D 128 PEs Fortran 2633.0 996.3 1885.4 NEC SX3/44 1 cpu 977.5 934.9 1831.0 Cray YMP/C90 2 cpu 869.1 759.7 1520.5 CM-5E 32 PEs 881.9 670.3 1339.4 Cray T3D 64 PEs assembly 1316.3 600.2 1168.4 Cray Y/MP 4 cpu 604.9 574.2 1154.3 CM-5E 32 PEs 734.9 516.0 1031.5 Cray T3D 64 PEs Fortran 1316.5 498.2 942.7 NEC SX3/12 1 cpu 479.4 416.2 832.3 NEC SX3/12 1 cpu 479.4 404.7 809.4 Cray YMP/C90 1 cpu 435.3 390.8 791.7 Cray T3D 32 PEs assembly 658.2 300.2 584.2 Cray T3D 32 PEs Fortran 658.3 249.1 471.4 CM-200 8k 10 MHz 328.2 247.1 471.1 Cray J94 4 cpu 297.1 182.7 362.2 ETA-10E 188.2 181.6 351.2 CM-2 8k 8 MHz 239.0 174.6 333.3 CM-2 8k 7 MHz 201.5 152.4 291.2 Cray Y/MP 151.6 143.9 283.1 Cray J94 2 cpu 163.1 102.6 202.7 Cray EL-98 8 cpu 144.4 98.9 197.0 Cray EL-98 4 cpu 98.1 80.6 163.0 Cray J94 1 cpu 87.2 52.6 106.2 Cray J94 1 cpu 85.4 52.5 101.0 Cray EL-98 2 cpu 52.1 43.7 89.9 FPS MCP728 7 cpu 64.8 37.1 74.0 Cray CS6400, 32 cpu 51.2 36.9 73.5 FPS MCP728 7 cpu 59.8 32.8 65.8 Cray CS6400, 24 cpu 47.1 32.3 64.5 SGI Power Challenge, 8 cpu 36.2 27.5 62.4 IBM RS6000-990 33.3 29.8 59.5 IBM RS6000-590 33.3 27.3 54.5 SGI Power Challenge, 6 cpu 33.7 25.1 53.9 Meiko CS-2 1 cpu 38.8 25.4 50.9 Cray CS6400, 16 cpu 37.6 24.8 49.6 Convex C3420 2 cpu 29.9 22.0 43.3 SGI Power Challenge, 4 cpu 25.5 19.3 39.9 Cray EL-98 1 cpu 27.3 22.3 39.7 Convex C3220 2 cpu 22.2 17.6 34.8 IBM RS6000-990/128 20.8 15.2 30.3 IBM 3090J/VF 1 cpu 20.8 14.4 29.0 Cray CS6400, 8 cpu 21.5 14.2 28.6 Convex C3220 2 cpu 20.8 14.1 28.0 IBM RS6000-580 17.9 15.6 27.8 Convex C3410 1 cpu 18.9 13.1 25.7 Convex C3420 2 cpu 17.1 12.6 25.0 Alliant FX/2800 14 cpu 19.1 12.4 24.3 DEC 7000/610 4 cpu 16.1 10.9 22.2 IBM 3090J/VF 1 cpu 15.6 12.0 22.1 SGI Power Challenge, 2 cpu 13.9 10.5 21.2 IBM RS6000-580 17.2 10.4 20.0 IBM RS6000-560 14.3 10.0 20.0 Cray T3D 1 PEs assembly 20.6 9.4 18.3 Convex C3210 1 cpu 11.5 9.1 18.2 IBM RS6000-550 15.0 8.2 18.0 Stardent P3 3 cpu 15.4 10.3 16.7 IBM RS6000-950 9.3 8.5 16.7 FPS 511 (1 vector) 11.3 9.0 16.4 IBM RS6000-950 11.1 8.2 16.0 FPS 511 SPARC vector 10.7 7.9 15.8 FPS 511 SPARC vector 10.7 7.9 15.8 Cray CS6400, 4 cpu 11.5 7.9 15.7 FPS 511 EA vector 10.5 7.8 15.6 FPS MCP104 1 cpu 13.0 9.7 15.5 FPS 511 EA vector 10.4 7.8 15.5 Convex C3210 1 cpu 11.2 7.6 15.1 Stardent ST2000 10.2 7.5 15.0 Convex C3410 1 cpu 10.8 7.5 14.9 Cray T3D 1 PEs Fortran 20.6 7.8 14.7 IBM RS6000-355 10.5 7.1 14.3 VAX 9000/420 1 cpu 10.3 7.2 13.1 VAX 9000/420 1 cpu 10.3 6.5 13.1 DEC 7000/610 2 cpu 10.4 6.7 13.1 DEC 4000/710 10.1 6.6 13.1 IBM RS6000-540 8.1 6.0 12.0 IBM 3090J scalar 7.0 6.0 11.4 SGI Pow Chal 8 banks 8.4 5.6 11.2 FPS MCP104 1 cpu 9.6 5.5 10.9 SGI Pow Chal 2 banks 8.2 5.4 10.8 Convex C1XP 7.8 5.4 10.8 DEC 2100 A500 2 cpu 8.3 5.4 10.6 SGI Challenge 150 MHz 8.3 5.6 10.5 IBM RS6000-530 5.5 5.1 10.0 IBM RS6000-355 8.3 4.8 10.0 Stardent Vistra 800b 9.2 4.8 9.6 IBM RS6000-530 6.7 4.5 9.5 IBM RS6000-320H 5.0 4.5 9.0 IBM RS/6000-250 6.4 4.4 8.9 SGI Crimson 6.7 4.5 8.7 IBM 3090J scalar 6.0 5.5 8.7 HP 9000/720 5.9 4.0 8.7 DEC 3000/500 6.0 4.1 8.3 SGI Challenge 1 cpu 5.9 4.2 8.0 IBM RS6000-320H 6.0 3.8 7.5 DEC 7000/610 1 cpu 5.7 3.7 7.2 DEC 4000/710 5.1 3.5 7.0 IBM RS6000-320 4.2 3.3 6.7 HP 9000/755 4.3 3.0 6.7 HP 9000/730 4.6 3.8 6.7 DEC 2100 A500 1 cpu 5.0 3.3 6.6 Alliant FX/80 8 cpu 4.5 3.2 6.4 Sun SS10/41 4.8 3.1 6.3 IBM RS/6000-250 4.2 3.1 6.1 Convex C1XP 4.4 3.0 5.9 Stardent P3 4.9 3.2 5.8 Sun SS10/30 5.2 3.2 5.7 SGI Challenge 150 MHz 3.6 2.4 5.2 SGI Crimson 3.7 2.4 5.0 IBM RS6000-320 3.8 2.5 5.0 Intel Pentium/60 3.9 2.6 4.9 MIPS RC6280 3.6 2.4 4.7 HP 9000/730 3.0 2.3 4.6 Sun SS10/41 1 cpu 3.0 2.2 4.5 SGI Indigo 2.6 2.2 4.5 SGI Challenge 1 cpu 3.5 2.2 4.5 Apollo DN10010 3.0 2.2 4.5 HP 9000/720 3.1 2.1 4.4 FPS 511 SPARC scalar 3.9 2.3 4.4 VAX 6000-410 vector 3.7 2.6 4.3 SGI 4D/35 3.6 3.1 4.2 Cray CS6400, 1 cpu 3.1 2.1 4.2 Alliant VFX80 3 cpu 3.0 2.1 4.2 Sun SS10/41 2 cpu 3.0 2.0 4.0 Sun 670 2.6 1.8 4.0 Sun SS10/30 2.9 1.9 3.9 Sun SparcClassic 3.0 2.0 3.6 Alliant VFX80 3 cpu 2.6 1.8 3.6 SGI 4D/35 2.3 2.0 3.5 Sun SS2 (4/75) 2.2 1.6 3.4 DEC 3000/300 2.1 1.7 3.2 DEC 5000/200 2.5 1.8 3.1 SGI Indigo 2.1 1.5 3.0 Sun SparcCenter 2000 2.3 1.5 2.9 FPS 511 SPARC scalar 2.8 1.5 2.9 Sun 4/490 1.8 1.9 2.7 SGI 4D/240 1.8 1.5 2.6 Sun 670 2.0 1.2 2.5 Sun SS2 (4/75) 1.9 1.2 2.2 Omron Luna88k 1.7 1.3 2.2 Sun 4/490 1.6 1.1 2.0 FPS 510 EA scalar 1.4 1.1 2.0 DEC 5000/200 1.5 1.0 1.9 FPS 510 EA scalar 1.4 1.0 1.8 SGI 4D/25 1.2 0.8 1.6 SGI 4D/240 1.0 0.8 1.6 Dell 486 DX/2-66 1.0 0.9 1.6 Sun SS1 1.1 1.0 1.4 NeXTStation 68040 1.0 0.9 1.4 NeXTStation 68040 0.7 0.7 1.3 Sun SS1 0.8 0.6 1.1 Omron Luna88k 0.9 0.7 1.1 SGI 4D/25 0.6 0.4 0.8