AMD processors played a key role in achieving a recent world record during the last simulation Ansys Fluent Computational Fluid Dynamics (CFD) at the Frontier supercomputer at Oak Ridge National Laboratory (Ornl). According to a press release AnsysHe ran 2.2 billion axial cells of the turbine simulation for Baker Hughes, an energy technology company, testing recent generation gas turbines to enhance performance. The simulation earlier took 38.5 hours to 3700 processor cores. Using 1 024 AMD MI250X accelerators in combination with AMD EPYC processors on the frontier, the simulation time shortened to 1.5 hours. It is over 25 times faster, which allows the company to see the impact of changes that it introduces on projects faster.
The recent supercomputer record has been set! Ansys, @bakerhughesco and @ornl have carried out the largest commercial #CFD simulation in history using 2.2 billion cells and 1024 GPU Instinct @AMD on the world’s first supercomputer exasca. Result? 96% reduction of simulation …April 4, 2025
Frontier was once the fastest supercompter in the world, as well as the first to break into the performance of Exascans. He replaced the Summit supercomputer, which was withdrawn from operation in November 2024. However, the supercomputer of El Capitan, located in the Lawrence Livermore national laboratory, broke the Frontier record at about the same time. Both Frontier and El Capitan are powered by AMD GPU, with the previous ones, 9,408 EPYC AMD processors and 37 632 AMD MI250X accelerators instincts. On the other hand, the latter uses 44 544 AMD Instinct MI300A accelerators.
Considering these numbers, the liquid CFD Ansys simulator apparently only used a fraction of power available on the border. This means that it can work even faster if it can exploit all available accelerators for a supercomputer. It also shows that despite the NVIDIA market dominance in AI GPU, AMD remains a powerful competitor, and its processors and GPU serve as the brains of some of the fastest supercomputers on Earth.
“Thanks to the scaling of CFD software with high loyalty to unprecedented levels with the power of AMD GPU processes, this cooperation shows how the most modern supercomputer can solve some of the most difficult engineering challenges, enabling a breakthrough in performance, balanced development and innovation,” said Brad McCredie Data engineering.
Although AMD can provide the highest quality performance at a much cheaper price than NVIDIA, many data centers and prefer Team Green due to software problems with AMD equipment.
One clamorous example was the TinyBox Tiny Corp system, which had problems with instability with AMD Radeon RX 7900 XTX graphics cards. The problem was so bad that Dr. Lisa Su had to enter to solve problems. And although it was allegedly repaired, the company still released two versions of TinyBox AI ACleler – one driven by AMD and the other by Nvidia. Tiny Corp also recommended the more exorbitant version of Team Green, with six RTX 4090 graphic processors, due to the quality of the driver.
If Team Red can repair software service on its great hardware, it can probably get more customers for their systems and get a more even position from NVIDIA on the AI GPU market.
