Running a memory-bound OpenFOAM case on an AMD EPYC cluster involves two practical decisions: how many nodes to request and how many MPI ranks to assign to each core complex. According to the paper's abstract, the fastest choice is frequently not the most economical one, so users face a trade-off between performance and cost.

The study, posted on arXiv, proposes a data-driven method to predict the optimal configuration. Because the abstract is truncated, the full methodology and results are not available from the source alone. The paper appears to revisit earlier work, but the details of that revisit are not described in the provided text.

This is a single-source digest, so there are no conflicting findings to compare. The main takeaway is that the paper addresses a practical scheduling problem for HPC users, where the best-performing setup may not align with budget constraints.