Molecular dynamics (MD) is an indispensable atomistic-scale computational tool widely-used in various disciplines. In the past decades, nearly all ab initio MD and machine-learning MD have been based on the general-purpose central/graphics processing units (CPU/GPU), which are well-known to suffer from their intrinsic “memory wall” and “power wall” bottlenecks. Consequently, nowadays MD calculations with ab initio accuracy are extremely time-consuming and power-consuming, imposing serious restrictions on the MD simulation size and duration. To solve this problem, here we propose a special-purpose MD processing unit (MDPU), which could reduce MD time and power consumption by about 10³ times (10⁹ times) compared to state-of-the-art machine-learning MD (ab initio MD) based on CPU/GPU, while keeping ab initio accuracy. With significantly-enhanced performance, the proposed MDPU may pave a way for the accurate atomistic-scale analysis of large-size and/or long-duration problems which were impossible/impractical to compute before.

分子动力学 （MD） 是各个学科中广泛使用的不可或缺的原子尺度计算工具。在过去的几十年里，几乎所有的 ab initio MD 和机器学习 MD 都基于通用的中央/图形处理单元 （CPU/GPU），众所周知，这些单元存在固有的“内存墙”和“电源墙”瓶颈。因此，如今具有 ab initio 精度的 MD 计算非常耗时和耗电，对 MD 仿真的大小和持续时间造成了严重限制。为了解决这个问题，我们在这里提出了一种专用的 MD 处理单元 （MDPU），与基于 CPU/GPU 的最先进的机器学习 MD （ab initio MD） 相比，它可以将 MD 时间和功耗减少约 10³ 倍（10⁹ 倍），同时保持 ab initio 的准确性。随着性能的显著提高，所提出的 MDPU 可能为以前不可能/不切实际地计算的大尺寸和/或长期问题的准确原子尺度分析铺平道路。