[lammps-users] Question about lammps running time

Dear LAMMPS users,

I am confused by the lammps running time. Now I am running lammps jobs on two different clusters, one is compiled with the Intel MPI compiler and the other is done with the native MPI compiler. I submitted the exact same jobs with same processors. However, for one case without KSPACE style I can get almost same efficiency, but for the case with KSPACE style pppm/tip4p, the Kspce time at the end of the log file run by the Intel MPI compiler cluster shows about five times longer than that of the native MPI compiler cluster.

Does any one know what’s the causes? Or other than the compiler, what else would influence the computing efficiency I should check?

Most appreciated.

Best
Yajie

Dear LAMMPS users,

I am confused by the lammps running time. Now I am running lammps jobs
on two different clusters, one is compiled with the Intel MPI compiler
and the other is done with the native MPI compiler. I submitted the
exact same jobs with same processors. However, for one case without
KSPACE style I can get almost same efficiency, but for the case with
KSPACE style pppm/tip4p, the Kspce time at the end of the log file run
by the Intel MPI compiler cluster shows about five times longer than
that of the native MPI compiler cluster.

in my experience, scaling issues rarely depend on
the compiler, but rather on hardware.

what kind of interconnect do the clusters have?
what kind of cpus/memory interface to the two clusters have?

cheers,
   axel.