GROMACS 5.1 vs 2016 performance in GPU-accelerated ceramide pull simulations

Computational
modeling of the skin barrier, the lipid matrix of the stratum
corneum, using molecular dynamics simulations. These studies give further insight about the largest organ in the
human body and will further clinical experiments. Thanks to the
highly optimized heterogeneous parallelization in GROMACS, complex
computational studies can be carried out quickly and efficiently. The
right panel shows the modeled molecular system with ceramide
molecules in green, cholesterols in white, and fatty acids in red.

Recent
work focused on improving SIMD, GPU, and thread
parallelization and resulted in speeding up the aforementioned
calculations
by
up to 50%. The left panel illustrates the execution time breakdown
of the CPU and GPU tasks in GROMACS versions 5.1 and 2016; improved
performance of multiple tasks leads to an increase in simulation
throughput from 61 ns/day to 95 ns/day.

The
simulations were performed on a workstation equipped with a Core
i7-5960X CPU and a GeForce TITAN X GPU.
Benchmarks and performance illustration (left panel) by S.P., the ceramide system model and rendering by M.L.