Hello, To make an efficient use of all available CPU cores I'd like to calculate many geometry points in parallel. If I understand properly this is a default behavior of cp2k when MPI-enabled build is started with mpirun. However, I've made a simple test with H2O molecule and discovered that for 8 parallel processes it makes 40 SCF calculations instead of 19 which are performed by cp2k.ssmp. And if I use just two processes it calculates 22 points. I wonder if this is a bug or an expected behavior, and how would cp2k behave if I run vibrational analysis for a very large system on, say, 192 CPUs? For now it seems to me like cp2k going to do a lot of unnecessary calculations, and it might be a better idea to stick with ssmp version and just do all points sequentially with OpenMP parallelism. My input file for H2O is attached. This is a full command that I used to run cp2k: numactl -N 0 --localalloc mpirun --use-hwthread-cpus --bind-to none --display-map -np 8 /home/enine/src/git/cp2k/exe/local/cp2k.popt -i H2O_b3lyp_freq.inp -o H2O_b3lyp_freq.out -- 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 visit <a href="https://groups.google.com/d/msgid/cp2k/020c8de2-bc5d-4385-b564-27110e90f941n%40googlegroups.com?utm_medium=email&utm_source=footer">https://groups.google.com/d/msgid/cp2k/020c8de2-bc5d-4385-b564-27110e90f941n%40googlegroups.com</a>.