The specific benchmark used was a molecular dynamics simulation which involved simulating a calcium aluminosilicate glass consisting of 6500 atoms at room temperature under atmospheric pressure for 25,000 timesteps, or 25ps. The radial cutoff distance was 5.5 Å. All of the data was able to fit into the physical RAM of each system, resulting in no swapping. However, the simulation code used tabulates the potentials and forces rather than recalculating them at each time step, making the performance depend heavily on memory performance (latency and bandwidth) and the cache hierarchy in addition to floating point arithmetic.

All binaries were compiled with optimizations for the platform on which the benchmark was being run, including any special SIMD instruction sets that the compiler could generate on its own. There are differences in the optimizations performed by the compilers, but I’ve tried to use each compiler’s most sensible set of optimization options. To see the exact options I used for each benchmark, in addition to a plethora of other metadata, you can view the raw data used to generate this page in my website’s git repository.

Note that some measurements have an “Optimized Time” column. This measurement was made on a version of the application that periodically sorts the atomic positions in memory to increase cache hit rate. Because this optimization was not available at the time I started doing these benchmarks, many of the older systems do not have it.

SPARC Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
Sun Ultra 10 UltraSPARC IIi 360 2048 KB EDO 5829 -
Sun Enterprise 4500 UltraSPARC II 400 4096 KB FPM 5639 -
Sun Netra T1 UltraSPARC IIi 440 2048 KB 5686 -
Sun Ultra 60 UltraSPARC II 360 4096 KB EDO 5561 -
Sun Ultra 60 UltraSPARC II 450 4096 KB EDO 4974 -
Sun Blade 100 UltraSPARC IIe 500 256 KB PC-133 8634 -
Sun Fire v100 UltraSPARC IIi 548 512 KB PC-133 5640 -
Sun Blade 150 UltraSPARC IIi 550 512 KB PC-133 5249 -
Sun Blade 150 UltraSPARC IIi 650 512 KB PC-133 5023 -
Sun Fire v210 UltraSPARC IIIi 1002 32+64 KB
1024 KB
DDR-266 2053 -
Sun Blade 1500 (Red) UltraSPARC IIIi 1062 32+64 KB
1024 KB
DDR-266 1862 -
Sun Fire v240 UltraSPARC IIIi 1503 32+64 KB
1024 KB
DDR-266 1558 -
Sun Blade 1500 (Silver) UltraSPARC IIIi 1503 32+64 KB
1024 KB
DDR-266 1501 -
Sun Blade 2500 (Silver) UltraSPARC IIIi 1600 1024 KB DDR2-266 1428 -

MIPS Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
SGI O2 MIPS R5000 180 512 KB - 25299 -
SGI O2 MIPS R10000 195 1024 KB - 14064 -
SGI Indigo2 MIPS R10000 195 1024 KB - 8438 -
SGI Octane MIPS R10000 225 1024 KB - 5744 -
SGI Fuel MIPS R14000 500 2048 KB - 2507 -
SGI Fuel MIPS R14000 600 4096 KB - 1787 -

POWER/PowerPC Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
IBM RS/6000 43p 150 PowerPC 604e 374 32+32 KB
1024 KB
PC-100 10889 -
Apple Power Mac G4 PowerPC 7450 733 256 KB PC-100 7291 -
Apple Power Mac G4 PowerPC 7450 933 2048 KB PC-100 4533 -
Apple Power Mac G5 PowerPC 970 1600 512 KB DDR2-400 2317 -
IBM Blue Gene/P PowerPC 450 850 32+32 KB
2 KB
8 MB
DDR2-840 3030 -
IBM Blue Gene/P PowerPC 450 4× 850 32+32 KB
2 KB
8 MB
DDR2-840 1720 -
IBM Power 550 Express POWER6 4204 64+64 KB
8 MB
32 MB
DDR2-667 646 -
IBM Power 550 Express POWER6 4× 4204 64+64 KB
8 MB
32 MB
DDR2-667 414 -

PA-RISC Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
HP C8000 PA-8800 1000 768+768 KB
32 MB
DDR2-266 801 -

Itanium Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
HP rx2600 Itanium 2 900 900 16+16 KB
256 KB
1.5 MB
DDR-266 1294 1212
HP rx2600 Itanium 2 900 2× 900 16+16 KB
256 KB
1.5 MB
DDR-266 1094 986

ARM Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
Raspberry Pi Model B ARM1176JZF-S 700 32+32 KB LPDDR2 16459 -
Raspberry Pi 3 ARM Cortex-A53 1200 32+32 KB LPDDR2 2531 -
Raspberry Pi 3 ARM Cortex-A53 2× 1200 32+32 KB LPDDR2 2427 -
Raspberry Pi 3 ARM Cortex-A53 4× 1200 32+32 KB LPDDR2 2508 -

x86 Processors

System Processor Clock (MHz) Cache Memory Time (sec) Optimized
Time (sec)
Dell Inspiron 8100 Pentium III M 1129 512 KB PC-133 3680 -
Tyan MPX S2466 Athlon MP 1900+ 1600 256 KB DDR-266 3660 -
Dell OptiPlex GX260 Pentium 4 (Northwood) 1793 512 KB DDR-333 2170 -
Dell OptiPlex GX260 Pentium 4 (Northwood) 2391 512 KB DDR-333 1661 -
Gateway E-4650 Pentium 4 (Northwood) 2193 512 KB PC-800 1743 -
Dell PowerEdge 2850 Xeon 2.8 (Nocona) 2793 1024 KB DDR2-400 974 -
- Pentium D 3200 2048 KB - 929 -
Asus Z70va Pentium M 760 2000 2048 KB - 731 -
Sun Fire X4100 Opteron 275 2200 1024 KB DDR-400 806 -
Sun Fire X2200 M2 Opteron 2214 2211 2048 KB DDR2-667 873 -
Sun Fire X2200 M2 Opteron 2376 2300 2048 KB DDR2-667 592 -
Supermicro Opteron 2216 2400 128+128 KB
1024 KB
DDR2-667 713 -
Supermicro X8DTT Xeon X5560 2800 8192 KB DDR3-1333 304 -
Dell Core 2 Duo E6700 2668 4096 KB DDR2-667 439 -
Dell OptiPlex 760 Core 2 Duo E7500 2926 3072 KB DDR2-667 346 -
Dell OptiPlex 990 Core i7-2600 3400 32+32 KB
256 KB
8MB
DDR3-1333 201 -
HP dc7800p Core 2 Duo E8400 2993 6144 KB DDR2-667 318 -
HP e9180f Core i7 920 2668 8192 KB DDR3-1333 334 -
HP ProLiant SL390s G7 Xeon X5672 3200 12 MB DDR3-1333 310 -
HP ProLiant DL360 G7 Xeon X5672 3200 12 MB DDR3-1333 308 -
- Xeon X5650 2668 12 MB DDR3-1333 316 -
- Xeon X5680 3334 12 MB DDR3-1333 313 -
- Xeon W5590 3334 8 MB DDR3-1333 273 -
Supermicro H8DGT Opteron 6212 2582 1× (64+16 KB)
1× 2 MB
1× 8 MB
DDR3-1066 353 -
- Opteron 6176 2300 64+64 KB
512 KB
6 MB
DDR3-1333 575 -
Supermicro SC818TQ-1400LPB Opteron 6136 2183 64+64 KB
512 KB
6 MB
DDR3-1333 565 483
Supermicro SC818TQ-1400LPB Opteron 6136 2× 2183 64+64 KB
512 KB
6 MB
DDR3-1333 477 419
Supermicro SC818TQ-1400LPB Opteron 6136 4× 2183 64+64 KB
512 KB
6 MB
DDR3-1333 412 407
Supermicro SC818TQ-1400LPB Opteron 6136 8× 2183 64+64 KB
512 KB
6 MB
DDR3-1333 476 470
Appro gB222X Xeon E5530 2400 32+32 KB
256 KB
8 MB
DDR3-800 416 -
Appro gB222X Xeon E5530 2× 2400 32+32 KB
256 KB
8 MB
DDR3-800 306 -
Appro gB222X Xeon E5530 4× 2400 32+32 KB
256 KB
8 MB
DDR3-800 265 -
Appro gB222X Xeon E5530 8× 2400 32+32 KB
256 KB
8 MB
DDR3-800 256 -
HP ProLiant SL230s G8 Xeon E5-2660 2989 32+32 KB
256 KB
20 MB
DDR3-1600 270 -
HP ProLiant SL230s G8 Xeon E5-2660 2× 2194 32+32 KB
256 KB
20 MB
DDR3-1600 208 -
HP ProLiant SL230s G8 Xeon E5-2660 4× 2194 32+32 KB
256 KB
20 MB
DDR3-1600 171 -
HP ProLiant SL230s G8 Xeon E5-2660 8× 2194 32+32 KB
256 KB
20 MB
DDR3-1600 171 -
HP ProLiant SL230s G8 Xeon E5-2660 16× 2194 32+32 KB
256 KB
20 MB
DDR3-1600 164 -
Appro gB512X Xeon E5-2670 2600 32+32 KB
256 KB
20 MB
DDR3-1333 292 237
Appro gB512X Xeon E5-2670 4× 2600 32+32 KB
256 KB
20 MB
DDR3-1333 175 155
Appro gB512X Xeon E5-2670 8× 2600 32+32 KB
256 KB
20 MB
DDR3-1333 164 141
Dell T5600 Xeon E5-2687W 3101 32+32 KB
256 KB
20 MB
DDR3-1600 205 -
Dell T5600 Xeon E5-2687W 2× 3101 32+32 KB
256 KB
20 MB
DDR3-1600 165 -
Dell T5600 Xeon E5-2687W 4× 3101 32+32 KB
256 KB
20 MB
DDR3-1600 133 -
Dell T5600 Xeon E5-2687W 8× 3101 32+32 KB
256 KB
20 MB
DDR3-1600 128 -
Dell PowerEdge C8220X Xeon Phi SE10P 8× 1095 32+32 KB
512 KB
GDDR5 1802 1209
Dell PowerEdge C8220X Xeon Phi SE10P 16× 1095 32+32 KB
512 KB
GDDR5 1589 -
Dell PowerEdge C8220X Xeon Phi SE10P 32× 1095 32+32 KB
512 KB
GDDR5 1479 1106
Cray XC30 Xeon E5-2695 v2 2401 32K+32K
256K
30M
DDR3-1866 279 225
Cray XC30 Xeon E5-2695 v2 2× 2401 32K+32K
256K
30M
DDR3-1866 216 173
Cray XC30 Xeon E5-2695 v2 4× 2401 32K+32K
256K
30M
DDR3-1866 180 163
Cray XC30 Xeon E5-2695 v2 6× 2401 32K+32K
256K
30M
DDR3-1866 176 154
Cray XC30 Xeon E5-2695 v2 12× 2401 32K+32K
256K
30M
DDR3-1866 163 149
Cray XC30 Xeon E5-2695 v2 12× 2401 32K+32K
256K
30M
DDR3-1866 166 153
Cray XC40 Xeon E5-2698 v3 2301 32K+32K
256K
40M
DDR4-2133 232 178
Cray XC40 Xeon E5-2698 v3 2× 2301 32K+32K
256K
40M
DDR4-2133 176 137
Cray XC40 Xeon E5-2698 v3 4× 2301 32K+32K
256K
40M
DDR4-2133 150 130
Cray XC40 Xeon E5-2698 v3 8× 2301 32K+32K
256K
40M
DDR4-2133 150 132
Cray XC40 Xeon E5-2698 v3 16× 2301 32K+32K
256K
40M
DDR4-2133 149 136
Cray XC40 Xeon E5-2698 v3 32× 2301 32K+32K
256K
40M
DDR4-2133 152 143
Cray XC40 Xeon Phi 7250 4× 1401 32K+32K
34M
MCDRAM 1259 839
Cray XC40 Xeon Phi 7250 8× 1401 32K+32K
34M
MCDRAM 1009 774
Cray XC40 Xeon Phi 7250 16× 1401 32K+32K
34M
MCDRAM 858 687
Cray XC40 Xeon Phi 7250 32× 1401 32K+32K
34M
MCDRAM 763 667
Cray XC40 Xeon Phi 7250 64× 1401 32K+32K
34M
MCDRAM 725 667
Cray XC40 Xeon Phi 7250 128× 1401 32K+32K
34M
MCDRAM 713 674
Cray XC40 Xeon Phi 7250 256× 1401 32K+32K
34M
MCDRAM 744 712
Apple iMac 18,3 Core i5-7600K 1× 3800 32K+32K
2×256K
6M
DDR-2400 189 143
Apple iMac 18,3 Core i5-7600K 2× 3800 32K+32K
2×256K
6M
DDR-2400 125 107
Apple iMac 18,3 Core i5-7600K 4× 3800 32K+32K
2×256K
6M
DDR-2400 115 97