My visit to ISC High Performance last month in Frankfurt, Germany re-affirmed my belief that computing innovation shows no signs of slowing down. I participated in an industrial HPC user panel at the event, which has traditionally focused on big supercomputing solutions for government and research institutions. The fact that this year’s ISC broke attendance records and dedicated so much time to industry sessions shows how much HPC has become entrenched in other industries.
We have been working with Intel on a few innovations that I wasn’t at liberty to discuss at ISC, but can now share with you that Intel announced its new processors and improvements to their accompanying technologies yesterday. We have been working with Intel to benchmark ANSYS software on the new technologies before their release, so that our mutual customers can immediately see what benefits they’ll receive. Here’s a sneak peek at the results.
1. Up to 59% faster performance of ANSYS Mechanical with the new Intel Xeon Gold 6148 processor.
To make it easier to differentiate its processors, Intel is now dividing them by scalability performance using metals (copper, silver, gold and platinum) to categorize them in Intel® Xeon® Scalable processors (formerly known by its code name, Skylake). Today’s processors have a number of capabilities beyond just processing, and the tiered series allows Intel to tailor its offerings based on data center client I/O, memory, storage and networking needs — with Copper at the entry level and Platinum at the top.
The Intel Xeon Gold 6148 processor, for instance, includes more cores, higher memory bandwidth, and an enhanced cache structure compared to the previous generation Intel Xeon processor E5 v4 family. To reap the full benefits of those enhancements, your engineering software must be capable of taking advantage of them.
Benchmark results show that a two-socket server based on the new Intel Xeon Gold 6148 processor can improve performance for ANSYS Mechanical by as much as 42% versus a previous generation server based on the Intel Xeon processor E5-2697 v4, and by as much as 59% versus a comparable server based on the earlier Intel Xeon processor E5-2698 v3 processor.
In Fluent, we’ve added support for Intel® Advanced Vector Extensions 2 (AVX2) optimized binary, so that we can take better advantage of the advanced vector processing capabilities of Intel Xeon processors. Our benchmark results also show the Intel Xeon Gold 6148 processor boosts performance for ANSYS Fluent 18.1 by up to 41% versus a previous-generation processor — and provides up to 34% higher performance per core.
Check out all the results in this white paper.
2. Up to 47% higher performance of ANSYS Fluent with Intel® Omni-Path Architecture vs. EDR InfiniBand
As Intel notes, computing performance depends on more than just processing speeds and core counts. Its Intel® Omni-Path Architecture (Intel® OPA) is a high-speed cluster fabric that helps avoid performance bottlenecks in clustered configurations. For example, Intel OPA enabled a 25% to 47% higher performance increase compared to the EDR InfiniBand networking communications standard across a range of ANSYS Fluent simulations run on large cluster configurations. This faster throughput will ultimately enable you as CFD user to simulate higher-fidelity models without having to expand your cluster nodes. For your CFD simulation, it basically improves the speed of communication among nodes on your current cluster, rather than throwing more hardware at the problem.
Read the white paper for all the ANSYS Intel OPA results.
3. Scalability for the biggest jobs via Knights Landing
The Intel® Xeon Phi™ processor is described as delivering “massive parallelism and vectorization to support the most demanding high-performance computing applications.” ANSYS has also been working with the new generation of Intel Xeon Phi, code named Knights Landing (KNL), and KNL with Intel OPA. We’ve seen good scalability for larger cases, but I must admit less so for smaller cases. Thanks to this research, our developers plan to look at ways to further improve vectorization in the Fluent solver code to be able to more benefit from the advanced vector processing capabilities of KNL.
As you can see, high-performance computing shows no signs of slowing down, and we continue to make sure that you can leverage Intel’s rapidly evolving platform to unlock new levels of performance and scaling with ANSYS software. Intel is a strategic partner for ANSYS, as together we expand the industry benefits of pairing high-performance computing with advanced simulation software.