| 10:54:40 AM PDT - Wed, Oct 21st 2015 |
|
TL;DR - what percentages of available memory should I allocate to stack and global for B3LYP/6-311G(d,p) optimizations/hessians and MP2/M06-L single point energies?
I mostly run optimizations of medium size systems (<100atoms) followed by a hessian at B3LYP/6-311G(d,p). I follow this with single point energy calculations with the same basis set but M06-L, B3LYP and MP2 theories. I use a few different clusters with different amounts of RAM per node. I already keep the heap memory down to 100mb/core, but am looking for some guidelines on how to allocate the rest. I've been using something like the default allocation for the rest (so about 1/3 for stack and 2/3 for global), but sometimes the hessians die because there isn't enough global memory available. Any guidelines?
|