Benchmark Results
| Software | Machine | OS | GPU | # | Polymer TPS | LJ liquid TPS |
|---|---|---|---|---|---|---|
| HOOMD 0.8.0 | Accelerator Cluster | Linux x86_64 | GTX 280 | 1 GPU | 247.9 | 248.8 |
| HOOMD 0.8.0 | NYX | Linux x86_64 | Tesla 1070 | 1 GPU | 217.5 | 225.2 |
| LAMMPS (12 Feb 2007) | lightning | Linux x86_64 | n/a | 32 CPU cores | 206.8 | 185.5 |
| HOOMD 0.8.0 | teslahoomd | Linux x86_64 | Tesla D870 | 1 GPU | 157.0 | 173.0 |
| HOOMD 0.8.0 | teslahoomd | Linux x86_64 | n/a | 1 CPU core | 6.73 | 6.87 |
Fields
- Software - The software used to perform the simulation
- Machine - Computer system the benchmark is performed on. See full hardware details of these systems below
- OS - Operating system software used
- GPU - The GPU HOOMD was executed on or n/a if not executing on a GPU
- # - Number of GPUs used in the benchmark (HOOMD) or the number of CPU cores used for non-GPU benchmarks.
- Polymer TPS - Performance of the polymer benchmark in Time steps Per Second
- LJ liquid TPS - Performance of the Lennard-Jones liquid benchmark in Time steps Per Second
Polymer system MD
The polymer system consists of of bead-spring ABA triblock polymers. 2,371 polymers of length 27 (for a total of 64017 particles) are simulated in implicit solvent at a packing fraction of 0.2. For more information on this polymer model, see the article 10.1027/ma061120f An average performance is measured over 50,000 time steps.
You can run this benchmark on your own machine by installing HOOMD and executing the benchmark script polymer_bmark.hoomd
The image shows a snapshot of a single time step in this benchmark with each sphere representing a single particle. Blue spheres are hydrophobic monomers and the orange ones are hydrophilic.
Lennard-Jones liquid MD
The Lennard-Jones liquid benchmark consists of 64,000 particles at a packing fraction of 0.2. An average performance is measured over 50,000 time steps.
You can run this benchmark on your own machine by installing HOOMD and executing the benchmark script lj_liquid_bmark.hoomd
The image shows a snapshot of a single time step in this benchmark with each sphere representing a single particle.
Notes
You can run these benchmarks yourself by installing HOOMD and executing the associated script. See the Installation Guide for more information on how to run benchmark.
VMD was used to create snapshots on this page and they were raytraced in tachyon.
Hardware details
Benchmarks are performed on various machines, ranging from the developers' office desktops to large clusters.
NYX @ The University of Michigan
Model: unknown
GPU: Tesla S1070 - 400 Series
RAM: 8 GiB DDR2
CPU: AMD Opteron 2218
Accelerator Cluster @ NCSA
Model: unknown
GPU: GTX 280
RAM: 4 GiB DDR2
CPU: AMD Opteron 2216
teslahoomd
Model: Sun Ultra 40 M2
GPU: Tesla D870
RAM: 2 GiB
CPU: 1 Dual-Core AMD Opteron(tm) Processor 2218
lightning
Model: Atipa cluster
148 nodes connected by InfiniBand, each with
RAM: 8 GiB
CPU: 2 Dual-Core AMD Opteron(tm) Processor 280
Click here for more details.