Ansys Fluent solvers using AMD Instinct GPUs reduce simulation run time by 96%, enabling more efficient high-fidelity CFD analysis.
Ansys announced notable results from a large-scale Fluent CFD simulation run on AMD Instinct MI250X GPUs. Using the Frontier exascale supercomputer with AMD EPYC CPUs and Instinct GPUs, Ansys and Baker Hughes scaled Fluent to 1,024 GPUs, providing detailed analysis of aerothermal physics at high operating pressures. By combining physical and virtual tests, Ansys simulation supports faster design cycles and cost-efficient development.
Traditional CFD methods involve lengthy development cycles and high costs for validating designs under extreme conditions. Exascale supercomputing systems supercharge computation, allowing for rapid iterations that shorten design-to-market timelines across applications. The combination of advanced hardware and leading multiphysics simulation software is pivotal for optimizing the development of turbine engines, power generation, and mechanical drives.
Baker Hughes uses Ansys Fluent to support the design of its next-generation gas turbines and other turbomachinery equipment to improve energy conversion efficiency and ultimately, reduce carbon footprints. Using the Frontier exascale supercomputer maintained by the Department of Energy’s Oak Ridge National Laboratory, Baker Hughes and Ansys ran a 2.2-billion-cell axial turbine stator simulation to identify critical flow and turbulence structures during the development phase.
When compared to methods using over 3,700 CPU cores, Baker Hughes and Ansys reduced simulation run time from 38.5 hours to 1.5 hours with 1,024 AMD Instinct MI250X GPUs. This highly scalable performance enables quicker design iterations, improves prediction accuracy, and supports the development of more sustainable technologies and products.
The advancements in the Fluent GPU solver also offer significant benefits to small and medium-sized businesses (SMB) operating on smaller GPU systems. With the solver’s improved computational efficiency and scalability, SMBs can achieve high-fidelity simulations without needing access to exascale resources.
For more information, visit ansys.com.
