Dear Farhat Tasnim ,

Does your machine have 8 CPU and 8 GB of memory? Admittedly it is very limited for the calculations you want to run and I would advise looking for something better.

Anyway, your only hope is to leverage OpenMP. Try running with

export OMP_NUM_THREADS = 8

mpirun -np 1 VASP

This should run VASP as a single process, split over the 8 cores with OpenMP. OpenMP should use better the memory at your disposal. Indeed, MPI replicates the same information on all the CPU involved in the calculation, while OpenMP shares the memory.

I am afraid, however, that your situation is dire and I am not expecting this solution to help you much.

Regards,

Roberto

Hi Farhat,

If you still have access to the old VASPv4.6 source code, you could compile the makeparam utility that came with older versions of VASP. That gives you a rough estimate of how much memory your job requires.

Also, what are your settings for KPAR, NPAR, etc? These influence the amount of memory required. If your job calculates all k-points in parallel, it will require much more memory than when you set all k-points the be calculated one after another.

Kind regards,

Peter

For production running, 8 core and 8 GB is fairly not enough.

If you just wanna learning, chose a small system less than 10 atoms with small kpoints and encut.