NAMD: Difference between revisions

Note: Using a verbs library is more efficient than using OpenMPI, hence only verbs versions are provided on systems where those are supported.  Currently verbs versions do not work on cedar as they are incompatible with the communications fabric, so use MPI version instead.

Newest '''NAMD 2.13''' is now also available.  To load the GPU-enabled versions, first run

:<code>module load cuda/10.0.130</code>

<translate>

This section shows an example of how you should conduct benchmarking of NAMD.  Performance of NAMD will be different for different systems you are simulating, depending especially on the number of atoms in the simulation.  Therefore, if you plan to spend a significant amount of time simulating a particular system, it would be very useful to conduct the kind of benchmarking shown below.  Collecting and providing this kind of data is also very useful if you are applying for a RAC award.

For a good benchmark, please vary the number of steps so that your system runs for a few minutes, and that timing information is collected in reasonable time intervals of at least a few seconds.  If your run is too short, you might see fluctuations in your timing results.   

The numbers below were obtained for the standard NAMD apoa1 benchmark.  The benchmarking was conducted on the graham cluster, which has CPU nodes with 32 cores and GPU nodes with 32 cores and 2 GPUs.  Performing the benchmark on other clusters will have to take account of the different structure of their nodes.

In the results shown in first table below we used NAMD 2.12 from verbs module. Efficiency is computed from  (time with 1 core) / (N * (time with N cores) ).

{| class="wikitable sortable"

|-

|}

These results show that for this system it is acceptable to use up to 256 cores.  Keep in mind that if you ask for more cores, your jobs will wait in the queue for a longer time, affecting your overall throughput.

Now we perform benchmarking with GPUs.  NAMD multicore module is used for simulations that fit within 1 node, and NAMD verbs-smp module is used for runs spanning nodes.

{| class="wikitable sortable"

|-

|}

From this table it is clear that there is no point at all in using more than 1 node for this system, since performance actually becomes worse if we use 2 or more nodes.  Using only 1 node, it is best to use 1GPU/16 core as that has the greatest efficiency, but also acceptable to use 2GPU/32core if you need to get your results quickly.  Since on graham GPU nodes your priority is charged the same for any job using up to 16 cores and 1 GPU, there is no benefit from running with 8 cores and 4 cores in this case.

Finally, you have to ask whether to run with or without GPUs for this simulation.  From our numbers we can see that using a full GPU node of graham (32 cores, 2 gpus) the job runs faster that it would on 4 non-GPU nodes of graham.  Since a GPU node of graham costs about two times what a non-GPU node costs, in this case it is more cost effective to run with GPUs.  So, you should run with GPUs if possible, however given that there are fewer GPU than CPU nodes, you may need to consider submitting non-GPU jobs if your wait for GPU jobs is too long.