MolPlay Benchmarks

Benchmarks were conducted using an automated script designed to extract hardware characteristics from a given machine and then run benchmarks for both the graphics card (GPU) and processor (CPU). The GPU benchmark was performed with UnityMol, while the CPU benchmark was conducted with Gromacs.

The extracted hardware characteristics in the table include:

• Number of CPU cores (#)
• CPU model and clock frequency
• Amount of memory (RAM)
• Type of GPU (N Nvidia or A AMD)
• Screen resolution used during the benchmark

Graphics performance is evaluated by rotating a large molecular structure—the 90S pre-ribosome (PDB-ID 6RXZ, 218,474 atoms)—360 degrees around its center using the HyperBalls graphics representation. The time taken to complete this operation is measured, with smaller values indicating better performance. A relative performance percentage is then calculated, comparing the results to a reference machine.

CPU performance is measured by simulating a fully hydrated DMPC lipid bilayer (16,853 atoms) for 20,000 steps to extrapolate the number of nanoseconds simulated per day. Higher values indicate better performance. As with the GPU, a relative performance percentage is calculated against a reference machine.

Benchmark entries are ranked according to relative CPU performance, as the CPU is often the limiting factor in most experiments.

Initial benchmarks on this site were run using the default (non-accelerated) graphics card driver. However, by configuring and tuning the graphics setup—following instructions available on the Batocera website—significant performance improvements can be achieved. For instance, on a test machine, a fivefold increase in performance was observed. Despite this, graphics performance is generally not the limiting factor in interactive simulations.

In the next MolPlay release, a benchmark routine will be available in the menu. This will generate a benchmark.csv file. If you’d like to contribute your results to our benchmark table, please send this file to us.