Hi everyone, I have a single node server with two Xeon Gold 6154 CPU @ 3.00GHz, 18 cores each, 36 total, 72 threads. I have tested several NPROC_REP values in a series of vibrational analysis runs started with the following command: mpirun -np 72 --bind-to hwthread cp2k.psmp -i cp2k.inp -o cp2k.out I report a brief table summary of the timings in the attached file: nproc_rep_test.txt The timings refer to the same two wf optimization steps, taken as example, for the elementary cell of a MOF of N=37 atoms. Each calculation started from the same cp2k-RESTART-1.wfn The required n. of SCF WAVEFUNCTION OPTIMIZATION is 223 (6N + 1). From the table, in principle, the best choice seems to be NPROC_REP=1. <div>Unfortunately, with NPROC_REP=1 each of the 72 replica will perform 4 SCF WAVEFUNCTION OPTIMIZATIONs, for a total of 288 calculations.</div><div>When replicas terminate the assigned calculations, the threads can not be redistributed towards the replicas that are still running. </div> So, these seem also aspects that need to be taken into account when choosing a suitable value of NPROC_REP. Please, let me know if you have any comments and thanks for your kind attention! Corrado -- You received this message because you are subscribed to the Google Groups "cp2k" group. To unsubscribe from this group and stop receiving emails from it, send an email to <a href="mailto:cp2k+unsubscribe@googlegroups.com">cp2k+unsubscribe@googlegroups.com</a>. To view this discussion on the web visit <a href="https://groups.google.com/d/msgid/cp2k/2dec963d-a086-4a0f-b539-23108a79db9bn%40googlegroups.com?utm_medium=email&utm_source=footer">https://groups.google.com/d/msgid/cp2k/2dec963d-a086-4a0f-b539-23108a79db9bn%40googlegroups.com</a>.