MOLECULAR DYNAMICS PERFORMANCE GUIDE - Digital Research Alliance of CANADA

BENCHMARK DETAILS

CPUY: CPU years per 1 microsecond long simulation. GPUY: GPU years per 1 microsecond long simulation. | T: tasks | C: cores | N: nodes. Speed is in ns/day. Integration step = 1 fs. Measured with dataset 6n40 (239,131 atoms).
ID Software Module Toolch Arch Data Speed CPU CPUeff CPUY GPUY T C N GPU NVLink Site
46 PMEMD.cuda.MPI amber/20.12-20.15 gofbc avx2 6n4o 8.71e+01 EPYC 7413 63.2 0.0 0.063 2 1 1 2A100-SXM4 Yes Narval
  • Benchmark submission script:
    1. #!/bin/bash

    2. #SBATCH --nodes=1 --ntasks=2 --gpus-per-node=a100:2

    3. #SBATCH --mem-per-cpu=2000 --time=1:0:0

    4. # Usage: sbatch $0

    5. 

    6. INPFILE=pmemd_prod.in

    7. STEPS=40000

    8. TMPFILE=tf_${SLURM_NTASKS}

    9. LOGFILE=production_${SLURM_NTASKS}.log

    10. module --force purge

    11. ml StdEnv/2020  gcc/9.3.0 cuda/11.4 openmpi/4.0.3 amber/20.12-20.15

    12. 

    13. # Print resource info

    14. echo ${SLURM_NODELIST} running on ${SLURM_NTASKS} tasks

    15. cat /proc/cpuinfo | grep "model name" | uniq

    16. 

    17. srun pmemd.cuda_SPFP.MPI -O -i $INPFILE -o $LOGFILE -p prmtop.parm7 -c restart.rst7

    18. grep "NonSetup CPU time" $LOGFILE > $TMPFILE

    19. 

    20. echo -n "ns/day:"

    21. awk  -v steps=$STEPS "{ print 3.6*2.4*steps*0.01/$6 }" $TMPFILE
  • Simulation input file:
    1. Benchmark

    2.  &cntrl

    3.   imin=0,irest=1,ntx=5,

    4.   nstlim=20000,dt=0.001,

    5.   ntc=2,ntf=2,

    6.   cut=8.0, 

    7.   ntpr=100, ntwx=1000,

    8.   ntb=2, ntp=1, taup=2.0,

    9.   ntt=3, gamma_ln=2.0, temp0=300.0,

    10.  /

    11.  &ewald

    12.   nfft1=128,

    13.   nfft2=128,

    14.   nfft3=128,

    15.  /