A Study on the Performance of Distributed Training of Data-driven CFD Simulations

2026-04-30

AI Summary

Traditional computational fluid dynamics (CFD) simulations are computationally expensive due to the need to solve partial differential equations, and while existing data-driven approaches accelerate inference, they still incur substantial training costs. This work proposes a deep learning-based time series model for efficiently predicting future fluid states and presents the first systematic evaluation of the trade-offs between training efficiency and prediction accuracy across CPU-only, multi-GPU, and distributed training strategies. Through minimal code modifications, we comprehensively analyze the acceleration and scalability of distributed training across diverse hardware configurations. Experimental results demonstrate that distributed GPU training significantly reduces training time while maintaining high prediction accuracy, achieving overall runtimes substantially lower than those of conventional CFD solvers.

Abstract

Data-driven methods for computer simulations are blooming in many scientific areas. The traditional approach to simulating physical behaviors relies on solving partial differential equations (PDE). Since calculating these iterative equations is highly both computationally demanding and time-consuming, data-driven methods leverage artificial intelligence (AI) techniques to alleviate that workload. Data-driven methods have to be trained in advance to provide their subsequent fast predictions, however, the cost of the training stage is non-negligible. This paper presents a predictive model for inferencing future states of a specific fluid simulation that serves as a use case for evaluating different training alternatives. Particularly, this study compares the performance of only CPU, multiGPU, and distributed approaches for training a time series forecasting deep learning (DL) model. With some slight code adaptations, results show and compare, in different implementations, the benefits of distributed GPU-enabled training for predicting high-accuracy states in a fraction of the time needed by the computational fluid dynamics (CFD) solver.

Problem

• distributed training
• data-driven CFD
• deep learning
• computational cost
• time series forecasting

Innovation

• distributed training
• data-driven CFD
• GPU acceleration
• deep learning
• time series forecasting