Arkansas High Performace Computing Center [hpcwiki]

User Tools

Site Tools

Trace:
namd2023

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
namd2023 [2024/02/28 21:04] – rootnamd2023 [2025/10/15 19:51] (current) – external edit 127.0.0.1
Line 34: Line 34:
 Single node 2.14 ``charmrun++ ++np 1`` with ``++ppn ##`` moved to left side should run equivalently to the same ``namd`` and same ``++ppn ##``.  Single node 2.14 ``charmrun++ ++np 1`` with ``++ppn ##`` moved to left side should run equivalently to the same ``namd`` and same ``++ppn ##``. 
  
-With two nodes, in a few cases ``charmrun++`` scales fairly well, but because of better alternatives, the prospects for worthwhile ``charmrun++`` runs are few.+With two nodes, in a few cases ``charmrun++`` scales fairly well, but because of better alternatives, the prospects for worthwhile ``charmrun++`` runs are few with this set of compute nodes.
  
 On Pinnacle I, 2.14 ``charmrun++ ++np 2`` scaled well but was still hardly faster than single-node 2.15a1 ``namd2``.  Three nodes didn't scale well at all, so there's not really a good use case for ``charmrun++``. On Pinnacle I, 2.14 ``charmrun++ ++np 2`` scaled well but was still hardly faster than single-node 2.15a1 ``namd2``.  Three nodes didn't scale well at all, so there's not really a good use case for ``charmrun++``.
Line 84: Line 84:
 ==GPU== ==GPU==
  
-Here using the number of cores available on the node (24/32/64) and one GPU (two or more GPUs ``devices 0,1,2,3`` scale poorly, not recommended or approved for AHPCC public use partitions).+Here we are using the number of CPU cores available on the node (24/32/64) and one GPU (two or more GPUs ``devices 0,1,2,3`` scale poorly, not recommended or approved for AHPCC public use partitions). This benchmark simulation scaled significantly with the CPU cores used up to the number of cores present.  A different test simulation didn't really scale at all and ``+p4`` was best.  It's not apparent to us from the input files why there is a difference. Test runs would be useful if you are going to do a lot of them.
  
-On the ``gpu72`` partitions with Intel 6130 and single NVidia V100, it's about 5 times faster than the best CPU version.  On ``agpu72`` partitions with AMD7543 and single A100, it's only about 10% faster than V100, so that's not a good use case for the more expensive AMD/A100 nodes.+On the ``gpu72`` nodes with Intel 6130 and single NVidia V100, it's about 5 times faster than the best CPU version, so are a good use case.  On ``agpu72`` nodes with AMD7543 and single A100, it's only about 10% faster than 6130/V100, so that's not a good use case for the more expensive AMD/A100 nodes, unless gpu memory requires the newer GPU.  The even more expensive multi-gpu ``qgpu72`` nodes also don't scale well over single-gpu and are not a good use case.
  
 <code> <code>
 +#gpu72/v100:
 module load namd/3.0a7 module load namd/3.0a7
 namd3 +p32 +setcpuaffinity +isomalloc_sync +devices 0 step7.2_production_colvar.inp namd3 +p32 +setcpuaffinity +isomalloc_sync +devices 0 step7.2_production_colvar.inp
 Info: Benchmark time: 32 CPUs 0.0393942 s/step 0.227976 days/ns 0 MB memory Info: Benchmark time: 32 CPUs 0.0393942 s/step 0.227976 days/ns 0 MB memory
- +#agpu72/a100: not recommended unless memory requires 
-namd3 +p64 +setcpuaffinity +isomalloc_sync step7.24_production.inp+namd3 +p64 +setcpuaffinity +isomalloc_sync _devices 0 step7.24_production.inp
 Info: Benchmark time: 64 CPUs 0.0344332 s/step 0.199266 days/ns 0 MB memory Info: Benchmark time: 64 CPUs 0.0344332 s/step 0.199266 days/ns 0 MB memory
 </code> </code>
  
namd2023.1709154283.txt.gz · Last modified: (external edit)