High Performance Cluster (HPC) Resource Allocation

Hi everyone,

recently I got access to our University HPC and started running MRtrix analyses on it. Since the HPC is using the slurm workload manager, I can (more or less freely) choose the amount of computational resources I need for the analyses. However, that’s the part that I am currently struggling with. I neither want to exaggerate and unnecessarily block resources, nor do I want to use too few resources and slow down the analyses.

So my question is: what would be the optimum value regarding RAM and CPU cores that maximizes the analysis speed, with regard to preprocessing, tractography (tckgen, 100mio streamlines), SIFT2 and connectome construction? And is there maybe even a maximum value for RAM and CPU cores, from which on MRtrix won’t get any faster?

Thank you very much for your help!

Best regards,