My Community

Dear Forum Members,
I need your expertise to optimize the efficiency of my calculations for a 4x4x4 supercell with 10 atoms in the unit cell (Zr and H). Can you please recommend me a good starting point for KPAR, NPAR and NCORE for running VASP compiled with GNU makefile without MPI. My computer has the following specifications:
• OS: Ubuntu 22.04 jammy
• Kernel: x86_64 Linux 5.15.0-76-generic
• Uptime: 5d 22h 50m
• Packages: 2723
• Shell: bash 5.1.16
• Disk: 16TB / 32TB (50%)
• CPU: Intel Xeon E5-2650 v4 @ 48x 2.9GHz [66.0°C]
• GPU: Quadro P4000
• RAM: 62368MiB / 64277MiB
We would be very grateful for your insights and best practices based on your experiences.
Best regards, Ingrid Švajger

Dear Ingrid Švajger,

I am afraid I do not understand your question. You ask for some rule of thumb on how to parallelise your jobs, but write that VASP was compiled without MPI support. How do you intend to achieve parallelisation this way?

Kind regards,
Pedro Melo

Dear Ingrid Švajger,

Just to add some information to my previous question, you can find an explanation about optimising the parallelisation strategy here wiki/index.php/Optimizing_the_parallelization.

Kind regards,
Pedro Melo

Dear Pedro Melo,
I'd like to thank you for your previous assistance in the parallelization of VASP. Thanks to your guidance and suggestion on MPI integration, I've successfully improved the parallelization of my jobs, which has led to better results. The link you provided to optimize parallelization was particularly useful.
Thanks and best regards, Ingrid Švajger