# Making Room for AI: Multi-GPU Molecular Dynamics with Deep Potentials in GROMACS

arXiv: https://arxiv.org/abs/2604.07276 · HTML: https://mlsanity.com/paper/2604.07276/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 46 references that are in the map.
- First posted on arXiv 8 Apr 2026.

## Builds on (most cited first, 18 of 18)

- [E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials](https://mlsanity.com/paper/2101.03164/) (2021)
- [MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields](https://mlsanity.com/paper/2206.07697/) (2022)
- [Learning Local Equivariant Representations for Large-Scale Atomistic Dynamics](https://mlsanity.com/paper/2204.05249/) (2022)
- [Machine learning for molecular simulation](https://mlsanity.com/paper/1911.02792/) (2019)
- [DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics](https://mlsanity.com/paper/1712.03641/) (2017)
- [TorchMD: A deep learning framework for molecular simulations](https://mlsanity.com/paper/2012.12106/) (2020)
- [A practical guide to machine learning interatomic potentials -- Status and future](https://mlsanity.com/paper/2503.09814/) (2025)
- [DPA-1: Pretraining of Attention-based Deep Potential Model for Molecular Simulation](https://mlsanity.com/paper/2208.08236/) (2022)
- [TorchMD-Net 2.0: Fast Neural Network Potentials for Molecular Simulations](https://mlsanity.com/paper/2402.17660/) (2024)
- [Extending the limit of molecular dynamics with ab initio accuracy to 10 billion atoms](https://mlsanity.com/paper/2201.01446/) (2022)
- [chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom MD simulations](https://mlsanity.com/paper/2506.04055/) (2025)
- [Heterogeneous Parallelization and Acceleration of Molecular Dynamics Simulations in GROMACS](https://mlsanity.com/paper/2006.09167/) (2020)
- [Towards Electronic Structure-Based Ab-Initio Molecular Dynamics Simulations with Hundreds of Millions of Atoms](https://mlsanity.com/paper/2104.08245/) (2021)
- [Scaling Molecular Dynamics with ab initio Accuracy to 149 Nanoseconds per Day](https://mlsanity.com/paper/2410.22867/) (2024)
- [Materials Learning Algorithms (MALA): Scalable Machine Learning for Electronic Structure Calculations in Large-Scale Atomistic Simulations](https://mlsanity.com/paper/2411.19617/) (2024)
- [MLMOD: Machine Learning Methods for Data-Driven Modeling in LAMMPS](https://mlsanity.com/paper/2107.14362/) (2021)
- [Redesigning GROMACS Halo Exchange: Improving Strong Scaling with GPU-initiated NVSHMEM](https://mlsanity.com/paper/2509.21527/) (2025)
- [Enabling AI Deep Potentials for Ab Initio-quality Molecular Dynamics Simulations in GROMACS](https://mlsanity.com/paper/2602.02234/) (2026)

## Abstract (arXiv)

GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.