Learn How to Gain Over 23x of Speedup in Discrete Element Modelling Simulation Leveraging NVIDIA CUDA Software for the Steel Manufacturing Industry

Discrete element methods (DEM) are any of a family of numerical methods for computing the motion and effect of a large number of small particles. DEM simulations are computationally intensive, which limits either the length of a simulation or the number of particles. Execution times can be reduced significantly with graphical processing units (GPUs). NVIDIA GPUs tend to be significantly more energy-efficient than conventional computing clusters when conducting DEM simulations; a DEM simulation solved on GPUs requires less energy and is multiple times faster than on a conventional computing cluster. We'll discuss a complex industrial-scale chute system (14-m high) used in the steel-making industry, simulated on the NVIDIA GP100 card using Altair EDEM software. For this application, we have reported speedups of over 23x with single precision and almost 18x with hybrid precision, compared with a 12-core CPU. New benchmarks are underway now using NVIDIA A100 cards, where a substantial performance improvement is expected.

Presented at NVIDIA GTC 2023 event by Carlos Labra, EDEM Product, Altair