simulation speed

Dear Lammps users,
I have simulated two systems one water+a slab of surface, another vacuum+slab of the same surface. Vacuum height is about 400 angstrom. The vacuum+surface is literally 4 times slower than the one with water despite the lesser number of molecules (1500 water molecules are replaced vacuum). Does anyone have any idea about this? sounds kind of wired for me …
Thanks you very much
Fatima

Dear Lammps users,
I have simulated two systems one water+a slab of surface, another vacuum+slab of the same surface. Vacuum height is about 400 angstrom. The vacuum+surface is literally 4 times slower than the one with water despite the lesser number of molecules (1500 water molecules are replaced vacuum). Does anyone have any idea about this? sounds kind of wired for me …

Not to me. I assume that you run in parallel? In that case you are likely to have a load balancing issue. Another potential reason would be using long range electrostatics. Kspace styles don’t exactly scale exactly with the number of atoms. In any case, it is difficult to state anything definite without seeing the exact inputs.

I would recommend to check out the processors keyword.

Axel

And look at the stats at the end of the run for the atoms/proc
histogram. If that is very imbalanced you will run more slowly.
So see the processors command, as Axel indicated.

Steve

Thank you very much Axel and Steve,
Yes the problem was bad allocations of the processors.

Best
Fatima