I am running polymer-water simulation. The polymer is modeled with OPLS-AA and water with TIP4P/Ice model. The system has approximately 33,000 atoms under NPT (1 bar, 275K) conditions with 10 angstrom cutoff for all non-bonded interactions.
I am using OMP acceleration with --ntasks=96 --cpus-per-task=1
.
Under such conditions, I see about 15 ns/day.
This seems somewhat slow to me. I have run simulation with 70k atoms and have seen 25ns/day. The disclaimer being that I have used intel acceleration. Now, LAMMPS does not support intel acceleration for l/cut/tip4p/long
interactions.
I wanted to ask if there is a way to speed up simulations any more under such conditions? Does TIP4P hold you to these speeds or is there a workaround?
These are my input files:
lammps input file: in.lammps (3.3 KB)
data file: pnipam_tip4p-ice-water.r1.data (5.7 MB)
settings file:pnipam_tip4p-ice-water.settings (6.6 KB)