Benchmarks

The benchmark results were produced by the scripts in examples/benchmarks, e.g.:

examples/benchmarks/generate_configuration.sh lennard_jones
examples/benchmarks/run_benchmark.sh lennard_jones

The Tesla GPUs had ECC enabled, no overclocking or other tweaking was done.

Simple Lennard-Jones fluid in 3 dimensions

Parameters:

  • 64,000 particles, number density \rho = 0.4\sigma^3
  • force: lennard_jones (r_c = 3\sigma, r_\text{skin} = 0.7\sigma)
  • integrator: verlet (NVE, \delta t^* = 0.002)
Hardware time per MD step and particle steps per second FP precision compilation details
Intel Xeon E5-2640 1.44 µs 10.8 double GCC 4.7.2, -O3
NVIDIA Tesla S1070 57.0 ns 274 double-single CUDA 5.5, -arch compute_12
  55.1 ns 284 single CUDA 5.5, -arch compute_12
NVIDIA Tesla C2050 39.4 ns 397 double-single CUDA 5.5, -arch compute_12
  34.3 ns 456 single CUDA 5.5, -arch compute_12
NVIDIA Tesla K20m 22.2 ns 702 double-single CUDA 5.5, -arch compute_12
  20.7 ns 756 single CUDA 5.5, -arch compute_12
NVIDIA Tesla K20Xm 19.7 ns 792 double-single CUDA 5.5, -arch compute_12
  18.4 ns 851 single CUDA 5.5, -arch compute_12

Results were obtained from 1 independent measurement based on pre-release version 1.0-alpha1. Each run consisted of NVT equilibration at T^*=1.2 over \Delta t^*=100 (10⁴ steps), followed by benchmarking 10⁴ NVE 5 times steps in a row.

Supercooled binary mixture (Kob-Andersen)

Parameters:

  • 256,000 particles, number density \rho = 1.2\sigma^3

  • force: lennard_jones with 2 particle species (80% A, 20% B)

    (\epsilon_{AA}=1, \epsilon_{AB}=.5, \epsilon_{BB}=1.5, \sigma_{AA}=1, \sigma_{AB}=.88, \sigma_{BB}=.8, r_c = 2.5\sigma, r_\text{skin} = 0.5\sigma)

  • integrator: verlet (NVE, \delta t^* = 0.001)

Hardware time per MD step and particle steps per second FP precision compilation details
Intel Xeon E5-2640 2.03 µs 1.93 double GCC 4.7.2, -O3
NVIDIA Tesla S1070 65.7 ns 59.4 double-single CUDA 5.5, -arch compute_12
  66.3 ns 58.9 single CUDA 5.5, -arch compute_12
NVIDIA Tesla C2050 39.2 ns 99.7 double-single CUDA 5.5, -arch compute_12
  32.4 ns 120 single CUDA 5.5, -arch compute_12
NVIDIA Tesla K20m 18.5 ns 211 double-single CUDA 5.5, -arch compute_12
  17.0 ns 230 single CUDA 5.5, -arch compute_12
NVIDIA Tesla K20Xm 16.2 ns 242 double-single CUDA 5.5, -arch compute_12
  15.0 ns 261 single CUDA 5.5, -arch compute_12

Results were obtained from 1 independent measurement and are based on pre-release version 1.0-alpha1. Each run consisted of NVT equilibration at T^*=0.7 over \Delta t^*=100 (2×10⁴ steps), followed by benchmarking 10⁴ NVE steps 5 times in a row.