Cray announced it has sold a CS500 cluster powered by AMD EPYC processors to the Haas F1 Team, which will use the system to run aerodynamic simulations for its Formula One racecars.
VF-18 racecar. Source: Haas F1 Team
“Every year we’re tasked with building the fastest and most efficient racecar in the most technologically advanced racing series in the world,” said Gary Foote, Chief Information Officer, Haas F1 Team. “We’re pleased to partner with Cray and utilize their HPC solutions. The double precision computing power of our new system will help solve our most demanding CFD and data challenges.”
Of course, if double precision capacity was the only factor, the Haas team might well have gone with an Intel Xeon chips, which can generally outrun EPYC processors in floating point performance. In truth though, Formula One sporting regulations limits computation fluid dynamics (CFD) performance used for wind tunnel simulation to 25 teraflops, so squeezing the best price-performance from these systems would appear to be much more of a driving factor.
In addition, the EPYC chips have demonstrated a particular affinity for CFD codes, which demand lots of memory bandwidth. As we pointed out in April when Cray announced it had added the AMD option to its CS500 line, independent testing showed that EPYC machinery ran CFD software significantly faster than its Xeon competition. That’s mainly because the EPYC chip has eight memory channels, compared to the six available on the latest Xeon “Skylake” CPUs.
Like other Formula One setups, the Haas F1 team will use the system to run wind tunnel simulations so that it can study the airflow patterns around its vehicles, enabling engineers to optimize their designs for better aerodynamics. In the Formula One business, less drag and more downforce are critical to building winning racecars.
Although most, if not all, Formula One engineers use HPC clusters nowadays to help design their racecars, few openly share information about their systems due to the intense competition between teams. The new Cray system is no exception. The supercomputer-maker didn’t offer much in the way of details on the Haas F1 cluster, other than its use of AMD’s EPYC 7000 processors and being paired with a Cray ClusterStor L300 storage system.
The system is scheduled for delivery in December 2018, in time for the 2019 racing season.