This bachelor’s thesis conducts a quantitative analysis to identify a potential replacement for Tetra Pak’s outdated hardware. The study assesses and compares the performance of the AMD EPYC 9334 processor, implemented in the C6 cluster node with the AMD EPYC 9174F processor, implemented in the F6 cluster node and the Intel Xeon Gold 6264, implemented in the C5 cluster node which is utilized as a basis for evaluation. The focus is on throughput, execution times, and optimal core counts for High Performance Computing cluster nodes. By evaluating these nodes using Abaqus, a Finite Element Analysis tool, on two implicit and one explicit simulation model, the research aims to determine the most efficient configuration for HPC cluster nodes through a series of benchmarks.

The study was divided into five distinct cases, numbered zero through four. Case zero presented the baseline values from each cluster node without modifications. Case one compared various Message Passing Interface configurations. Case two tested uncommon core counts. Case three compared Local Scratch and General Parallel File System and case four presented co-scheduling. Results gathered signified the importance of utilizing the Message Passing Interface configuration in conjunction with co-scheduling to enhance the performance and reduce execution times. Following each test case, calculation tables were created the present the price differences between the C6 and F6 cluster nodes, determining the most cost-efficient solution. From each test case and calculation, F6 presents itself as the best choice overall due to license costs where F6 performed better per number of cores.