This depends on the supercomputer hardware (cores per node, mabye GPUs) and the NAMD build (MPI?, SMP?), so giving you general instructions is difficult without knowing anything about the supercomputer you want to run your jobs on.

Assuming a normal network enabled SMP build of NAMD, you'd use one process per node which might be "ntasks" and threading this process across the remaining cores "cpus-per-task". However, please realize that this script you might have, is not a commonly known standard, rather than a template for that particular supercomputing facility, so without us seeing it, we can't tell for sure.

Dear Norman Geist

What about MPI? I think Bib Alex has 24 cores per node.

Norman Geist The NAMD build is MPI and the supercomputer is Bib Alex.

Do you need the script file I'm using, sir?

Yes seeing that script file will help

Norman Geist These are the lines in the script file I'm using.

#!/bin/sh

#SBATCH --job-name=namd-sample300-1

#SBATCH --partition=cpu

#SBATCH --ntasks=300

#SBATCH --cpus-per-task=1

#SBATCH --time=00-50:20:00

namd2 step4_equilibration.conf > num300_1.log

I tried to change ntasks and cpus-per-task because it takes nearly 100 minutes for only 50 picoseconds for ubiquitin's minimization step.

If you have an MPI build of NAMD, than you need to start it via mpirun. Write mpirun in front of the namd2 command. Than you need to select the number cpus for the simulation. If a node has 24 cores, than use multiples of 24 for --ntasks.