Dear lammps users,
I have the exactly same data file and input file, which is a 10000 atoms polymer structure. I ran the script both on my desktop and on a cluster.
My desk top use AMD quad-cure cpu and have 4GB memory. On the cluster I use 1 node with two hex-core cpu with 12GB memory. but the simulation result is totally different.
I know there will be some differences if we ran simulation on different computers, but those results are totally different.
also, I don’t think my desktop have any problem, because I ran a 20000 atoms polymer simulation on it and it gives the same(or say similar) result as I run on that cluster.
Could someone please give some explanations? How to solve such problem or something I should be beware of?
Thank you all
Yueqi
Please clarify what constitutes a totally different result.
Axel.
Are you using same version of LAMMPS?
Oh, thanks for point this out. I do used a different version of lammps.
On my desktop, I use the LAMMPS windows parallel executable – C++ version (1 July 2012)
On the cluster, they build the LAMMPS 31Aug11 with the linux redhat version.
I will try to run the script on my desktop under the Ubuntu system again and see how it goes.
Oh, thanks for point this out. I do used a different version of lammps.
but that is not the reason. the reason for the fast divergence between
the two trajectories lies in the use of the langevin thermostat. this
contains a random component obtained from a pseudo random number
generator and that *cannot* be consistent when you run with a
different number of MPI processes, even if you use the same random
seed.
in any case, there is no indication that the trajectories give a
"different result". the actual individual coordinates of the total
system and kinetic energy are hugely irrelevant in MD simulations.
what matters are the averages over the ensemble and - provided you are
in equilibrium - over time. this seems to be given here and thus the
simulations are *not* giving a different result. i suggest to read up
on this in a text book on MD simulations, as this is one of the
fundamental concepts here and must be understood to be able to use MD
successfully.
axel.