JADE2 Technical Notes

AMBER install

• The MPI/cuda installation on JADE2 is a bit scuffed, so it's not possible to compile AMBER22, so AMBER20 was used for all benchmarks (via via module load amber/20).

GROMACS

• Used the module gromacs/2022.2

NAMD

• NAMD 3.0 alpha 7 was used, as it provided on JADE2 via module load namd/3.0-alpha7
• The script file needs an extra line to support GPUs, that line is: CUDASOAintegrate on

OpenMM

• The provided module is scuffed, so OpenMM was installed via conda. The most recent version is 8.1 as of january 2024 (when these tests were done). To correctly install conda, you need to use compatible gcc and cuda versions from conda-forge:

conda install -c conda-forge gcc=12.1.0
conda install -c conda-forge cudatoolkit=11.4
conda install -c conda-forge openmm=8.1.0

LAMMPS

• Load gcc 9.1.0 and cuda/11.1-gcc-9.1.0
• Build from source with cmake -D PKG_GPU=on -D PKG_MOLECULE=on -D PKG_KSPACE=on -D PKG_RIGID=on -D PKG_EXTRA-DUMP=on -D GPU_API=cuda -D GPU_ARCH=sm_70 -D PKG_OPT=on ../cmake/