Comparative Study of CPU vs. GPU Acceleration in Molecular Dynamics Simulations

EasyChair Preprint 14904

Abstract

Molecular dynamics (MD) simulations are essential tools for studying the behavior of molecules and materials at an atomic scale. Traditionally, MD simulations have relied on central processing units (CPUs) for computation, but the growing complexity of molecular systems and the need for higher resolution and longer time scales have prompted the adoption of graphics processing units (GPUs) to accelerate these calculations. This comparative study evaluates the performance of CPU vs. GPU acceleration in MD simulations, focusing on computational efficiency, accuracy, and scalability. Using benchmark molecular systems, we assess the speedup factors, energy efficiency, and the potential trade-offs in terms of model precision and resource utilization. The results show that GPU acceleration offers significant performance improvements, particularly for large systems, without compromising accuracy. However, certain limitations in memory management and workload distribution across CPUs and GPUs are identified, which could impact scalability for extremely large simulations. This study highlights the evolving role of GPUs in enhancing MD simulation workflows, offering insights into their optimal application and future directions for hybrid computing architectures.

Keyphrases: CPU vs. GPU Acceleration, Molecular dynamics (MD) simulations, computational efficiency

@booklet{EasyChair:14904,
  author    = {Adaan Ahsun},
  title     = {Comparative Study of CPU vs. GPU Acceleration in Molecular Dynamics Simulations},
  howpublished = {EasyChair Preprint 14904},
  year      = {EasyChair, 2024}}