[gmx-users] nvidia tesla p100
irem.altan at duke.edu
Thu Nov 3 17:04:56 CET 2016
I think I solved the problem. The system that I’m testing might be just too small that there aren’t enough non-bonded interactions to increase the load on the GPU sufficiently to result in a performance increase. When I enlarged the system to add (10 times more) solvent molecules, running the simulation on a GPU was 2-3x faster than running it on the CPU.
More information about the gromacs.org_gmx-users