An update on namd for the shared memory one-node version namd2/namd3
. At this writing the production version of NAMD is 3.0b5 for CPU and GPU, which requires Rocky 8 so can't run on our production compute nodes. With current GPUs the standard NAMD benchmark apoa1 is too small to show the actual performance, so here we use a user's lipid simulation for 25k steps.
Examples follow for the newest versions we can run. These don't have modules as all they need is a PATH and they will soon be outdated.
Using the number of cores available on the node, either “+p32” or “+p64”.
export PATH=$PATH:/share/apps/NAMD/./NAMD_2.13_Linux-x86_64-multicore namd2 +p32 +setcpuaffinity +isomalloc_sync step7.2_production_colvar.inp
Again using the number of cores available on the node (24/32/64) and one GPU (two or more GPUs devices 0,1
scale poorly, not recommended or approved for public use partitions).
export PATH=$PATH:/share/apps/NAMD/NAMD_3.0alpha7_Linux-x86_64-multicore-CUDA namd3 +p32 +setcpuaffinity +isomalloc_sync +devices 0 step7.2_production_colvar.inp
Partition Cores Proc GPU Used Walltime _____________________________________ comp72 32c i6130 0 5979 acomp72 64c a7543 0 2282 tres72 32c a6136 0 running gpu72 32c i6130 1xV100 1 1168 pcon06 32c i6130 2XV100 2 1038 agpu72 64c a7543 1xA100 1 884 tgpu72 24c i4166 4xTitanV 1 1638 tgpu72 24c i4166 4xTitanV 2 1472