The declining utility of Generalized Born implicit solvent models over the past decade owes to shortcomings in accuracy and poor scaling to modern computer architectures. In contrast, vectorized computation in general purpose graphics processing units (gpGPUs) promotes representations of all solvent atoms (TIP3P, SPC/E) for the newfound ability to simulate many particles on timescales relevant to enzymatic binding and catalysis. Implicit solvent models retain advantages for small biopolymer folding problems and ligand docking studies, where solvent friction inhibits motions of interest. A new implementation of Generalized Born molecular dynamics (GBMD) in the mdgx program of AmberTools stages hundreds of problems on a single gpGPU, in effect turning it into a miniature beowulf cluster. The amount of L1 cache available on gpGPU multiprocessors limits each simulation to 928 atoms or less. Drastic reductions in global memory transactions and the number of threads applied to any one problem help the implementation engage up to 80% of the total instruction throughput of the device. A comparable GBMD in the Amber pmemd program that applies the whole gpGPU to a single problem shows severe parallel scaling bottlenecks for systems smaller than 2500 atoms, but the mdgx implementation can run many 250-atom problems as fast as the pmemd implementation can run just one. For problems in the applicable range, the mdgx implementation improves simulation throughput by a factor of ten (for 900 atom problems) to a factor of 100 (for very small, 100 atom problems) on an NVIDIA V100 gpGPU. One gpGPU can simulate multiple replicas of a single protein or a variety of different proteins in the same run. The new implementation is expanded for debut in Amber20 to include replica exchange and nudged elastic band capabilities. We apply the new capabilities to study protein folding pathways and outline extensions to bring scalable GBMD energetics in line with explicit solvent models.

Harnessing gpGPUs for multi-replica molecular simulations with implicit solvent

Macromolecular Chemistry: The Second Century

[Author Benefits](https://scimeetings.acs.org/?utm_source=pubs_content_marketing&utm_medium=website&utm_campaign=0320_MCF_NPI_Launch_Spring_Homepage&ref=pubs_content_marketing)

[How to use SciMeetings](https://storage.googleapis.com/dl.morressier.com/acs/SciMeetings-ACSSpring2020-HowToGuide_M3020-Final.pdf)

[Frequently Asked Questions](https://storage.googleapis.com/dl.morressier.com/acs/SciMtng_ACSNMSp2020_FAQ_M3120_Final.pdf)

ACS Spring 2020 National Meeting & Expo

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.