Hi all,
I have a large system with 64 cores (4 socket) and I was wondering if anyone has experience with the best way to run various parts of Relion on this kind of system. We have 512GB of RAM, so I assume that for most processes running 64 threads would be the best. However, 3D Autorefine requires MPI jobs, and I was wondering how they will make use of my available cores. If I set it to 3 MPI processes, will each take the same number of cores? Once I have a second GPU the choice will probably be clear, but for now does anyone have any advice?
Best regards,
Kasim Sader
|