How many processors can I use with KPAR Paralellization?
Posted: Thu Oct 17, 2013 5:02 pm
Hello all,
I am doing optical properties calculations on a super cell of ~60 atoms, and need to use a very dense kpoint mesh to accurately capture the optical properties.
I have been using KPAR to increase efficiency but I still have not converged the optical properties of my supercell convincingly, and my next step requires 1200 kpoints. Using the scaling of kpar and nprocs that I have been this next job will be with KPAR=6 and nprocs = 768.
Does this kind of massive parallelization affect my VASP output? And should I try and reduce my memory requirements elsewhere? Or can VASP effectively handle being parallelized over 700+ processors?
I am doing optical properties calculations on a super cell of ~60 atoms, and need to use a very dense kpoint mesh to accurately capture the optical properties.
I have been using KPAR to increase efficiency but I still have not converged the optical properties of my supercell convincingly, and my next step requires 1200 kpoints. Using the scaling of kpar and nprocs that I have been this next job will be with KPAR=6 and nprocs = 768.
Does this kind of massive parallelization affect my VASP output? And should I try and reduce my memory requirements elsewhere? Or can VASP effectively handle being parallelized over 700+ processors?