Gamers may have been the first to enjoy the benefits of GPU technology. Today it has grown into many areas of enterprise and supercomputing where it packs an impressive punch.
Now a team led by Peter Vincent, a senior lecturer in the Department of Aeronautics at Imperial College London, is giving CFD technology a 10x boost with the launch of the new GPU-accelerated PyFR v1.0.0.
Computational Fluid Dynamics is the computational simulation of fluid flow. It helps engineers understand complex air and fluid flow patterns without building a wind tunnel. They’re incredibly helpful when designing aircraft, wind turbines and even F1 racing cars.
Meet the Next Generation
PyFR is an open-source CFD package that employs new, very accurate methods known as Flux Reconstruction schemes to tackle some of the world’s most challenging fluid flow problems. In particular, problems where the flow is unsteady and turbulent.
PyFR works on a range of hardware platforms, including large-scale clusters powered by the NVIDIA Tesla Accelerated Computing Platform.
Why use Tesla GPU accelerators? Their many-core architecture is the key. While the average modern laptop has two to four cores, the NVIDIA Tesla GPU accelerators has many more.
Vincent is using more than 2,500 ultra-high performance compute cores. The new CFD software runs on clusters of hundreds of these GPU accelerators. So these systems can make billions calculations every minute, processing vast amounts of data.
Accelerating the Future
Early results have generated excitement in the CFD community. A study presented at the AIAA Aviation conference in Dallas this week indicates that GPU-accelerated PyFR can achieve 10x increases in accuracy and efficiency.
Aerospace industry leader BAE Systems has been a close collaborator. “We regard PyFR’s numerical accuracy, algorithmic and parallel performance as a benchmark for other codes,” said Oscar Neilson from BAE Systems UK. World-class aerodynamics has been key to the UK’s success in the aerospace sector. It plays a vital role in the aerodynamic design of aircraft, allowing engineers to understand complex airflow patterns without ever flying a real aircraft.
It’s a big deal as well for someone like H.T. Huynh from NASA’s Glenn Research Center, which aims to advance our exploration of the solar system. “To me, PyFR is a game changer,” said Huynh. “It has caused me to re-evaluate my research as well as my path forward. I hope to make extensive use of this remarkable and elegant program.”
Whether engineers will use PyFR to develop the next F1 winners or airplanes that can take us into space and back, this is yet another industry being turbo-charged by the power of GPU technology.