On the recently-held (June 24-July 2) International Supercomputing Conference (ISC), Intel showcased how the corporate goals to increase its lead in excessive efficiency computing (HPC) with a variety of expertise disclosures, partnerships and buyer adoptions. The chip big introduced advances in its Xeon processor for HPC and AI in addition to improvements in reminiscence, software program, exascale-class storage, and networking applied sciences for a variety of HPC use circumstances.
“To maximize HPC performance we must leverage all the computer resources and technology advancements available to us,” mentioned Trish Damkroger, vp and basic supervisor of Excessive Efficiency Computing at Intel. “Intel is the driving force behind the industry’s move toward exascale computing, and the advancements we’re delivering with our CPUs, XPUs, oneAPI Toolkits, exascale-class DAOS storage, and high-speed networking are pushing us closer toward that realization.”
Earlier this yr, Intel strengthened its place in HPC with the launch of third Gen Intel Xeon Scalable processors. The newest processor delivers as much as 53% greater efficiency throughout a variety of HPC workloads, together with life sciences, monetary companies and manufacturing, as in comparison with the earlier technology processor.
In comparison with its closest x86 competitor, the third Gen Intel Xeon Scalable processor is claimed to ship higher efficiency throughout a variety of well-liked HPC workloads. Xeon Scalable 8380 processors are additionally claimed to outperform AMD EPYC 7763 processors on key AI workloads, with 50% higher efficiency throughout 20 frequent benchmarks.
HPC labs, supercomputing facilities, universities and authentic tools producers who’ve adopted Intel’s newest compute platform embody Dell Applied sciences, HPE, Korea Meteorological Administration, Lenovo, Max Planck Computing and Knowledge Facility, Oracle, Osaka College and the College of Tokyo.
The corporate claims that its next-generation of Xeon Scalable processors (code-named “Sapphire Rapids) will provide built-in Excessive Bandwidth Reminiscence (HBM), offering a dramatic enhance in reminiscence bandwidth and a major efficiency enchancment for HPC purposes that function reminiscence bandwidth-sensitive workloads. Customers can energy by means of workloads utilizing simply Excessive Bandwidth Reminiscence or together with DDR5.
The early main wins for Sapphire Rapids processors with built-in HBM embody the US Division of Power’s Aurora supercomputer at Argonne Nationwide Laboratory and the Crossroads supercomputer at Los Alamos Nationwide Laboratory.
The Sapphire Rapids-based platform will present capabilities to speed up HPC, together with elevated I/O bandwidth with PCI categorical 5.0 (in comparison with PCI categorical 4.0) and Compute Specific Hyperlink (CXL) 1.1 assist, enabling superior use circumstances throughout compute, networking and storage.
Along with reminiscence and I/O developments, Sapphire Rapids is optimized for HPC and synthetic intelligence (AI) workloads, with a brand new built-in AI acceleration engine referred to as Intel Superior Matrix Extensions (AMX).
Extending Intel Ethernet for HPC
At ISC 2021, Intel additionally introduced its new Excessive Efficiency Networking with Ethernet (HPN) answer, which extends Ethernet expertise capabilities for smaller clusters within the HPC section by utilizing normal Intel Ethernet 800 Sequence Community Adapters and Controllers, switches based mostly on Intel Tofino P4-programmable Ethernet swap ASICs and the Intel Ethernet Cloth suite software program. HPN is claimed to allow software efficiency corresponding to InfiniBand at a decrease price whereas benefiting from the convenience of use supplied by Ethernet.
Business assist for DAOS
Intel rolled out business assist for DAOS (distributed asynchronous object storage), an open-source software-defined object retailer constructed to optimize information change throughout Intel HPC architectures. DAOS is on the basis of the Intel Exascale storage stack, beforehand introduced by Argonne Nationwide Laboratory, and is being utilized by Intel prospects reminiscent of LRZ and JINR (Joint Institute for Nuclear Analysis).
DAOS assist is now obtainable to companions as an L3 assist providing, which permits companions to offer a whole turnkey storage answer by combining it with their companies. Along with Intel’s personal information heart constructing blocks, early companions for this new business assist contains HPE, Lenovo, Supermicro, Brightskies, Croit, Nettrix, Quanta, and RSC Group.