Atomistic molecular dynamics (MD) simulations are a vital tool in chemical research, as they are able to provide a view of chemical systems and processes that is not obtainable through experiment. However, large-scale MD simulations require access to multicore clusters or supercomputers that are not always available to all researchers. Recently, many have begun to explore the power of graphics processing units (GPUs) for various applications, such as MD. We present preliminary results of water simulations carried out on GPUs. We compare the performance gained using a GPU versus the same simulation on a single CPU or multiple CPUs. We also address the use of more accurate double precision arithmetic with the newest GPUs and its cost in performance. Key words: Molecular dynamics, GPU, CUDA
Joseph E. Davis, Adnan Ozsoy, Sandeep Patel, Miche