In context: Computational fluid dynamics (CFD) exploits the facility of pc expertise to check and predict how fluids behave below simulated circumstances. Supercomputers can do wonders with CFD purposes, and AMD Intuition GPUs have been correctly leveraged to scale back the time wanted to run these simulations, with room for much more enchancment.
Ansys, a simulation software program firm, used its Fluent software program to simulate a fancy turbine system on one of many world’s strongest supercomputers. By working the duty on AMD Intuition GPUs as an alternative of conventional CPU cores, the corporate minimize simulation time by 96 %. Ansys says that is only the start of dashing up computational fluid dynamics workloads.
Ansys Fluent offers superior physics modeling to research numerous fluid phenomena. The corporate just lately partnered with power agency Baker Hughes, which is designing next-generation fuel generators and different turbomachinery to spice up power conversion effectivity. Collectively, they labored with Oak Ridge Nationwide Laboratory and its Frontier supercomputer – the primary HPC system to achieve exascale-class efficiency.
Frontier runs on AMD Epyc CPUs and Intuition GPUs, which have gained traction at the same time as Nvidia continues to semi-monopolize the info middle GPU market. Ansys and Baker Hughes tapped into the supercomputer’s monumental processing energy to scale its Fluent simulation throughout 1,024 GPUs. The end result: a 2.2-billion-cell axial turbine stator simulation that helped pinpoint key stream and turbulence constructions early within the design part.
Ansys credited AMD’s Intuition GPUs for slicing the simulation time from 38.5 hours to simply 1.5. The sooner technique relied on greater than 3,700 CPU cores, however the brand new method ran on 1,024 Intuition MI250X GPUs. Based on the corporate, this marks a brand new scaling document for CFD purposes. This breakthrough might dramatically pace up design iterations and ship extra correct efficiency forecasts for industrial programs.
Ansys additionally sees potential advantages for small and mid-sized companies. Even with out entry to large HPC programs, corporations utilizing smaller GPU-based setups might nonetheless see important good points in simulation effectivity and scalability. Ansys claims enterprises can speed up complicated CFD workloads with no need large-scale infrastructure by working Fluent on GPU-based programs.
For now, Baker Hughes is utilizing the Ansys Fluent GPU solver and Frontier supercomputer to optimize the event of its next-generation turbine engines. Ansys believes combining fashionable GPUs and superior CFD software program may benefit different complicated design and industrial purposes.
