How to Enable Complex Simulations: The Power of Multiphysics & Digital Thread Seminar

Exploring GPU Acceleration in Computer-Aided Engineering: Advantages, Innovations, and Prospects

Author: Ian Pegler - NVIDIA

Abstract

This presentation focuses on the practical use of Graphical Processing Units (GPUs) to accelerate Computer-Aided Engineering (CAE) tasks. GPUs are known for their ability to perform numerous tasks simultaneously, making them particularly suitable for distributing computational workloads in CAE simulations. The presentation delves into real-world scenarios where GPU acceleration proves advantageous in CAE, specifically in areas like structural mechanics, fluid dynamics, and electromagnetic simulations. These examples illustrate how GPU-powered simulations provide quicker insights into product behavior.

Additionally, the presentation introduces a new CPU-GPU architecture named Grace-Hopper. This architecture aims to overcome memory bandwidth limitations, potentially enhancing performance in applications that are only partially accelerated. By discussing Grace Hopper, the presentation offers a glimpse into future advancements in hardware that could further amplify CAE capabilities. In addition, a short introduction to NVIDIA’s framework for physics AI models, Modulus, will also be presented.

In summary, this presentation provides an objective exploration of GPU utilization in expediting CAE computations. It presents practical cases, discusses the potential of the Grace Hopper architecture, and emphasizes the tangible benefits and challenges of integrating GPU acceleration into the field of engineering simulation.