| Summary: | NAMD benchmarks were done on five different proteins with varying system sizes (anoplin, kalata B1, North-Atlantic ocean pout antifreeze protein, Pseudomonas aeruginosa PAO1 lipase and octopamine receptor in mushroom bodies, OAMB) solvated with TIP3P water through four different publicly available computer resources in the Philippines. Our results show that the high-end desktop generated the most ns/day for small and medium-sized systems (e.g. anoplin, kalata B1, and antifreeze protein) while BlueGene/P generated the most ns/day for larger system sizes (e.g. lipase and octopamine receptor). Although these computing resources are capable of exploring protein behavior through molecular dynamics (MD) simulations for small to medium-sized systems, dealing with large systems require tremendous computational resources. This benchmark highlights the importance of intercommunication in NAMD. Moreover, our results showed the advantage of using GPU-accelerated desktops for certain MD simulations. However, the poor scalability of the high-end desktop does not make it viable for simulating large systems. Improvements in Philippine computing infrastructure and protocol is highly recommended to keep up with advances in high performance computing globally.
|
|---|