Last week NVIDIA released its tweak of the MD code AMBER, optimized to perform well on its GPU platform (AMBER is part of NVIDIA’s recently announced Bio Workbench)
AMBER 11, the latest version of one of the most widely used applications for biochemists and others involved in molecular dynamics research, is optimized to run on NVIDIA® graphics processing units (GPUs), which speed up the tool by up to 100-fold over a traditional CPU -based server. GPUs deliver performance from a desktop workstation that previously could only be achieved on a supercomputer, improving productivity as scientists no longer need to wait for time on a shared supercomputer or departmental cluster of servers.
…AMBER 11 is designed to take advantage of NVIDIA Tesla™ 20-series GPUs, which utilize the massively parallel CUDA™ architecture for the specific needs of high performance computing applications. In early trials with the AMBER user community, Dr. Walker received over a dozen reports of speedups over 30-times on a range of bio-molecular simulations.
…Dr. Walker noted that just one Tesla GPU provides as much processing power as a high performance cluster of 512 CPU cores when simulating one nanosecond in the lifetime of a 25,000 atom implicit solvent nucleosome. In a simpler explicit solvent simulation, such as the JAC benchmark, he noted that one Tesla GPU provides the equivalent processing power of 48 cores of the NSF Ranger supercomputer.