COSMO-SMD optimization of large systems


Click here for full thread
Clicked A Few Times
I am trying to do implicit solvent of a fairly large organic molecule (184 atoms, 896 electrons) using NWChem. I am wondering if there are some studies on what would be a better ("ideal") memory and processor requirements for such a job.

1. NWChem's benchmark shows very good scaling over 10^3 procressors, however I wasn't able to see any speed up between 56 and 112.

2. Is there to know if the program is memory constrained? I have tried with 32gb and 64Gb memory and I can't see any difference in the speed.

Each solvent phase optimization step of COSMO-SMD is currently taking around an hour (even at sto-3g basis set) which seems excruciatingly slow!

Any help will be greatly appreciated!