This guest post explores Intel’s advanced webinar series that focuses on understanding HPC fabrics and Intel Omni-Path Architecture.
To help prospective customers understand Intel OPA capabilities and inform new customers of how to take advantage of all that Intel OPA has to offer, Intel began presenting a series of webinars on the fabric nearly three years ago.
High Performance Computing system architects choose Intel Omni-Path Architecture (Intel OPA), Intel’s high-speed fabric, for its performance and price/performance, driven by its 100 Gbps wire rate with very high message injection rates and low latency. With Intel OPA’s 48-port radix switch silicon, architects can build simpler networks that deliver full bi-sectional bandwidth, while using less hardware for lower networking costs. That means HPC customers, like Spain’s Barcelona Supercomputing Center (BSC), Japan’s Joint Center for Advanced HPC (JCHPC), and Texas Advanced Computing Center (TACC), can spend more of their budget on compute power.
Intel OPA’s success in the market is shown by how it continues to lead in the rankings among 100 Gbps fabrics in the Top500 list. The Top500 is a good proxy for the rest of the marketplace. Between the June 2017 and November 2017 lists alone, Intel OPA increased 20 percent above the previous six months in total FLOPS and listings of the top 100 systems. Some of the customers leading among the top 100 systems include:
- #7 Oakforest-PACS at Joint Center for Advanced High Performance Computing (JCAHPC) in Japan
- #12 Stampede 2 at the Texas Advanced Computing Center (TACC) at University of Texas in the U.S.
- #13 MareNostrum at Barcelona Supercomputing Center (BSC) in Spain
- #14 Marconi at CINECA in Italy
The use of HPC has extended beyond traditional simulation and analytics. We’re seeing the acceleration of deep learning algorithms on HPC clusters through multi-node processing. Customers who have now deployed clusters with the new Intel Xeon Scalable processors and Intel OPA, such as BSC, are finding that the new processor along with Intel’s fabric create a great platform for deep learning training across multiple nodes.
To help prospective customers understand Intel OPA capabilities and inform new customers of how to take advantage of all that Intel OPA has to offer, Intel began presenting a series of webinars on the fabric nearly three years ago. Since March of 2015, Intel has broadcast 28 webinars, with new ones coming approximately monthly. These webinars cover a wide range of topics, from the important factors to consider when selecting or designing HPC networks to Intel OPA implementations and the benefits seen by customers to Intel OPA in new workloads, such as AI. After each interactive webinar session, the program is available on-demand at Intel’s Webinar site for HPC. Some recent programs include:
Accelerating AI with Intel Omni-Path Architecture
Intel OPA Webinar Series Part 27
Learn how Intel OPA plays a role in streamlining and reducing the time to train machine learning frameworks.
Intel Omni-Path Architecture: Advanced
Intel OPA Webinar Series Part 25
Get insights from Intel’s director of high performance fabric marketing on the features in Intel OPA that will take your HPC clusters to the next level of performance. Learn how Intel OPA provides accelerated interconnectivity with a variety of CPU/GPCPU, storage or other environmental options.
Bringing Up an Intel Omni-Path Architecture-based Cluster in a University Environment
Intel OPA Webinar Series Part 20:
Learn how the Texas Tech High Performance Computing Center is taking maximum advantage of Intel OPA in their newest cluster.
Stay tuned for more webinars. We are in the final planning stages of yet another informative session. To keep informed of upcoming webinars and other Intel OPA news, subscribe to the Intel Omni-Path Adviser.
