Reference Guide

Altair Radioss Performance
19 Dell EMC Ready Solutions for HPC Digital Manufacturing with AMD EPYC™ Processors—Altair Performance
Here, the results are significantly different than the results obtained with the hybrid parallel version of
AcuSolve. For the larger Taurus (T) model, the best performance was always obtained using a single shared
memory thread (the same as non-hybrid distributed memory MPI). For the smaller Neon model, there was a
small benefit using more than one thread at four or more nodes. These results indicate that users should
carefully test their models for the potential benefit of using multiple threads when running on larger number of
nodes, particularly with smaller models.