Arkansas High Performace Computing Center [hpcwiki]

User Tools

Site Tools

Trace: • gromacs2023
gromacs2023

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision		Previous revision
gromacs2023 [2024/01/20 23:55] – rootgromacs2023 [2025/10/15 19:51] (current) – external edit 127.0.0.1
Line 1: Line 1:
-=== Gromacs 2023.2===+=== Gromacs 2023.2/2023.3===
  
-Several versions of the [[ http://www.gromacs.org ]] molecular dynamics program are installed. The most complete and latest is version 2023.2.  Only the single precision version is installed at this time.  The various computers are compared with the GMX50 water benchmark. As shown, the GPU version is much faster. Because of poor scaling over multiple GPUs, Gromacs is not approved for multiple GPU public-use servers. +Several versions of the [[ http://www.gromacs.org ]] molecular dynamics program are installed. The most complete and latest is version 2023.2.  Only the single precision version is installed at this time.  The various computers are compared with the GMX50 bare water benchmark [[https://ftp.gromacs.org/pub/benchmarks/]] . As shown below, the GPU version is much faster. Because of poor scaling over multiple GPUs, Gromacs is not approved on multiple GPU public-use servers. 
-These results reflect four separate installations that are autoselected by the module at runtime. Multiple node runs are not yet tested and will require a machinefile as shown in [[mpi|mpi]]. These versions are compiled with OpenMP but OMP_NUM_THREADS>1 on CPU-only nodes did not improve results. +These results reflect four separate installations that are autoselected by the module at runtime. Multiple node runs are not yet tested and will require a ``machinefile`` and possibly ``-genv`` as shown in [[mpi|mpi]]. These versions are compiled with OpenMP but OMP\_NUM\_THREADS>1 on CPU-only nodes did not improve results. ``gromacs/2023.3`` may also be selected in both CPU and GPU.  Performance appears similar. CPU version with gcc or GPU version with icx have not been compiled
      
 <code> <code>
Line 8: Line 8:
 Node Type                     GPU       NS/day  partition Node Type                     GPU       NS/day  partition
  
-32-core dual Intel 6130       none       2.645  comp+32-core dual Intel 6130       none       3.021  comp
 32-core dual Intel 6130       1xV100    16.242  gpu 32-core dual Intel 6130       1xV100    16.242  gpu
 64-core dual AMD 7543         none       7.057  acomp 64-core dual AMD 7543         none       7.057  acomp
Line 15: Line 15:
 With $NC=32 or 64 physical cores: With $NC=32 or 64 physical cores:
 CPU: CPU:
 +
 module load intel/23.0.0 python/intel-3.7.7 impi/21.8.0 mkl/22.1.0 gromacs/2023.2 module load intel/23.0.0 python/intel-3.7.7 impi/21.8.0 mkl/22.1.0 gromacs/2023.2
 OMP_NUM_THREADS=1 mpirun -np $NC gmx_mpi mdrun -resethway -npme 0 -notunepme -noconfout -nsteps 1000 -v -s  bench.tpr OMP_NUM_THREADS=1 mpirun -np $NC gmx_mpi mdrun -resethway -npme 0 -notunepme -noconfout -nsteps 1000 -v -s  bench.tpr
  
 Single GPU: Single GPU:
-nvidia-smi -c 0  #shared mode#+ 
 +sudo nvidia-smi -c 0  #set shared mode#
 module load gcc/11.2.1 cuda/12.1 impi/19.0.9 python/intel-3.7.7 gromacs/2023.2 module load gcc/11.2.1 cuda/12.1 impi/19.0.9 python/intel-3.7.7 gromacs/2023.2
 OMP_NUM_THREADS=$NC gmx_mpi mdrun -nb gpu -resethway -npme 0 -notunepme -noconfout -nsteps 1000 -v -s  bench.tpr OMP_NUM_THREADS=$NC gmx_mpi mdrun -nb gpu -resethway -npme 0 -notunepme -noconfout -nsteps 1000 -v -s  bench.tpr
 </code> </code>
gromacs2023.1705794929.txt.gz · Last modified: (external edit)