Gmx_mpi vs gmx

[shivam1]

what difference (performance wise) it makes if I execute any related command with gmx only rather using gmx_mpi?

I am using 32 core CPU with 1 NVIDIA GPU. As per my understanding, MPI library is for parallel threading and as I am not using multiple GPUs so for me use of gmx and gmx_mpi is the same.

Please correct me if I misunderstood.

However I am not able to execute command using gmx_mpi.

Thanks

[plagueis]

gmx_mpi only applies if you have GROMACS installed with real MPI as opposed to thread-MPI

Single node performance should be marginally better with thread-MPI. The only case in which you would want to use real MPI on a single node is when running multiple simulations that require real-time communication - i.e. multiple walker metadynamics or other replex algorithms - as these don’t work with thread-MPI.

[shivam1]

@plagueis

Single node performance should be marginally better with thread-MPI. The only case in which you would want to use real MPI on a single node is when running multiple simulations that require real-time communication - i.e. multiple walker metadynamics or other replex algorithms - as these don’t work with thread-MPI.

Thanks @plagueis
Sorryy, This made me more confused. I just wanted to know for the above configuration what should I use (for fast simulation) gmx or gma_mpi?

and why am I not able to use gmx_mpi while I installed gromacs with mpi on
Thanks