Performance comparison of CFD-DEM solver MFiX-Exa, on GPUs and CPUs

TL;DR

This paper compares the computational performance of a CFD-DEM solver on GPUs and CPUs, demonstrating significant speed-ups and efficiency gains with GPU use and adaptive time stepping in large-scale gas-solid simulations.

Contribution

It provides a detailed performance analysis of MFiX Exa on GPUs versus CPUs, highlighting the benefits of hybrid parallelism and adaptive time stepping for large-scale simulations.

Findings

Single GPU is 10x faster than a CPU core.

3 GPUs outperform 64 CPU cores by 4x.

Adaptive time stepping yields 4x speed-up on both architectures.

Abstract

We present computational performance comparisons of gas-solid simulations performed on current CPU and GPU architectures using MFiX Exa, a CFD-DEM solver that leverages hybrid CPU+GPU parallelism. A representative fluidized bed simulation with varying particle numbers from 2 to 67 million is used to compare serial and parallel performance. A single GPU was observed to be about 10 times faster compared to a single CPU core. The use of 3 GPUs on a single compute node was observed to be 4x faster than using all 64 CPU cores. We also observed that using an error controlled adaptive time stepping scheme for particle advance provided a consistent 4x speed-up on both CPUs and GPUs. Weak scaling results indicate superior parallel efficiencies when using GPUs compared to CPUs for the problem sizes studied in this work.