More than 20 percent of computing time on the world’s 500 fastest supercomputers goes to modeling atomic and molecular motion. Such work supports material design, drug interaction studies and protein folding research.
Using a standard mathematical function, scientists at the Simons Foundation’s Flatiron Institute created a technique that runs these simulations 2.5 to seven times faster without loss of accuracy. For the popular GROMACS software package the method produced a fivefold speedup at high precision.
The approach integrates readily into existing workflows, promising lower time and energy costs. The findings were published online May 21 in Nature Communications.
Senior author Shidong Jiang noted that many scientific fields relying on molecular simulations will benefit from reduced energy and computing demands. Nvidia’s Anthony Costa called the work a significant advance after years of only small gains.
Molecular dynamics models track millions of molecules over tiny time steps, sometimes requiring trillions of calculations. The new method addresses the heavy cost of computing long-range electrostatic forces between charged particles.
