My Community

Dear all,
I compiled vasp 5.2.11 similar to version 5.2, but now the runtime went slow dramatically. It is runing on 12 cores like version 5.2 on 1 core. Does anybody know the reason for this behavior of the code ?
buudo

Please provide some more data to have a better comparison than "slower", e.g. version, runtime, number of cores, compiler version etc.

Thanks,

Alex

Assuming you calculate on a 12-core-per-socket machine, are you sure you used the latest MPI libraries? At least Intel MPI 3.x fails to recognize the 12 cores per socket, and identifies 8 physical cores and 4 'virtual' cores - which leads to an unfavorable 12 threads on 8 cores. No problems observed with MPI 4.0.