Home > Workload Solutions > High Performance Computing > White Papers > Dell Validated Design for HPC Digital Manufacturing with Altair Simulation Suite and 3rd Generation Intel Xeon > Multiserver scalability
The following figure shows the parallel performance for the Taurus benchmark model using up to eight servers equipped with dual Intel Xeon 8358 32-core processors using from one, four and eight threads per rank. Radioss supports hybrid parallel operation and in some instances, this may provide better performance than MPI only. The number of ranks per server was equal to the number of cores per server divided by the number of threads per rank.
The larger values shown in Figure 4 represent better performance. Reference performance is obtained from running the benchmark on a single server using one thread per MPI task. For this benchmark model, using one thread per MPI process provides the best performance and scalability. However, this behavior is problem dependent, so users are encouraged to test various threading options with their specific models to determine the best runtime settings.