White Papers

Extreme GPU Computing
Page 11
4.3 Accelerating Molecular Dynamics with NAMD
Figure 10: NAMD performance and acceleration compared to CPU-only
The advent of hardware accelerators has influenced Molecular Dynamics by reducing the time to results
and therefore providing a tremendous boost in simulation capacity. LAMMPS and GROMACS are two
open source Molecular Dynamics (MD) applications that can take advantage of these hardware
accelerators, and have seen a tremendous boost in simulation capacity. NAMD is a freely available,
feature-rich, GPU-enabled molecular dynamics simulator.
In this section, we evaluate NAMD performance improvement wit, on two proteins F1ATPASE and STMV.
These proteins consist of 327K and 1066K atoms respectively. The performance measure is in
“days/ns”, which shows the number of days required to simulate one nanosecond of real-time.
Figure 10 shows the performance of NAMD on the PowerEdge C4130. Configurations A, B and C are the
four GPU configurations. However the acceleration on NAMD also seems to be sensitive to number of
CPUs, e.g., there is a significant difference in the acceleration between A and B. B has an additional
CPU compared to A. Among the three 4 GPU configurations the current version of NAMD performs best
on configuration C. The difference in two highest performing configurations C and B is the manner in
which GPUs connect to the CPUs. The balanced configuration C has two GPUs attached to two CPUs
resulting in 7.8X acceleration over the CPU-only runs. The same four GPUs attached via a switch
module to a single CPU results in about 7.7X acceleration. On the dual GPU configurations, D performs
better with 5.9X acceleration compared to 4.4X in configuration E. Configuration D performs better,
which is in line with the assumption that a 2nd CPU is helpful for NAMD, as we observed B & C (the dual
CPU quad GPU configurations) performing best among the four GPU configurations.
Four K80 Boards
Two K80 Boards