I am simulating a bead-spring model of a polymer in LAMMPS with a FENE interaction between bonded atoms and Lennard-Jones interaction between non-bonded atoms. I am running this on a GPU cluster. Here is my input script.
############################################################
LAMMPS POLYMER MOLECULAR DYNAMICS
#Bead spring model of polymer with FENE and LJ interaction
############################################################
package gpu 1
Box and units (use LJ units and periodic boundaries)
units lj # lennard-jones units
atom_style bond # atoms with bonds
boundary p p p # all boundaries are periodic
read_data initial_configuration.txt
FENE potential and LJ interaction
Between bonded atoms
bond_style fene
special_bonds fene # Prevents LJ from being counted twice
bond_coeff 1 10.0 1.5 0.305 1.0
Between non-bonded atoms
pair_style lj/cut/gpu 2.5
pair_modify shift yes
pair_coeff 1 1 1.0 1.0 1.122461
pair_coeff 1 2 1.0 1.0 1.122461
pair_coeff 2 2 2.0 1.0 2.5
Set up fixes
variable seed equal 54654651
fix 1 all nve # NVE integrator
fix 2 all langevin 1.0 1.0 1.0 ${seed} # langevin thermostat
Output thermodynamic info (temperature, energy, pressure, etc.)##########################
thermo 1000
thermo_style custom step temp etotal pe ke epair emol press vol
set timestep of integrator
timestep 0.005
run
dump 1 all atom 10000 dump.lammpstrj
dump 2 all custom 10000 dump.position.lammpsbin id type x y z vx vy vz
run 1000000
########## Input script end here #########
There are two types of beads in the polymer. Number of beads is 512, Number of bonds is 511.
My simulation works without using GPU, but when I attempt to simulate it on the GPU, it runs initially for 12000 steps only and then stops with the following error.
Cuda driver error 700 in call at file ‘/lfs/sware/lammps/lammps-2Aug2023/lib/gpu/geryon/nvd_timer.h’ in line 76.
Abort(-1) on node 1 (rank 1 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, -1) - process 1
I tried changing the time-step also but it stops exactly after 12000 steps.
I am running this using 4cpu and 1gpu node.