Never miss a story from Computational Fluid Dynamics. Subscribe for in-depth analysis and articles.
The Intel® Xeon® processor family has long been a popular choice for CFD hardware, both in workstations and in HPC clusters. The Intel® Xeon® processor Scalable family in particular is targeted at HPC applications, where it has been the computation workhorse in government, academia, and industry. We finally got to take some of our Fidelity CFD technology for a spin on the new fourth-generation Intel® Xeon® Scalable Processors, formerly called Sapphire Rapids. There are two major upgrades over the previous, third-generation line (formerly Ice Lake). Intel has moved to a new process node, Intel 7, and a tile-based architecture, which enables them to put more cores on the package than ever before. However, as CFD users and developers know, CFD computations tend to be bandwidth-bound, and more cores don't do you much good if you don't have the memory bandwidth to keep them fed with data. Last generation, we saw that eight channels of DDR4-3200 were enough for up to around 32 cores, but not much more than that. The fourth-generation Intel® Xeon® Scalable Processors have up to eight channels of DDR5-4800, delivering a 1.5X bandwidth improvement over the third generation, so CFD users can make the most of these higher core counts.
We provided Intel with a benchmark using our density-based solver technology on the classic Windsor Body. Windsor is a classic, public geometry that features many of the important features of a van or SUV, and since it has experimental results extant, it is a workhorse case here at Cadence for our CFD validation studies. We've run HPC scaling studies in the past with Windsor body, so it's also a great test for checking out high-performance CPUs like these fourth-generation Intel® Xeon® Scalable Processors.
Figure 1: RANS simulation of the Windsor body, which was tested on fourth-generation Intel® Xeon® Scalable Processors
A RANS simulation run on fourth generation Intel® Xeon® Scalable Processors demonstrates the performance benefits over third generation. With higher bandwidth, higher core counts, and a generational IPC uplift, all thanks to the new process node and architecture, the fourth generation is almost twice as fast as the third generation when running Fidelity CFD software. CFD engineers know that you can never have enough compute power, and with this new generation of Xeon Scalable hardware, Intel has delivered what engineers need. More power and more compute, which means higher-accuracy results in less time than ever before.
Figure 2: Speedup of Fidelity CFD software various fourth generation Intel® Xeon® Scalable Processors processors over the third-generation 8360Y
Disclaimer: Intel, the Intel logo, and Xeon are trademarks of Intel Corporation or its subsidiaries.