Intel introduced this week that its processors, computational accelerators and sturdy Optane DC reminiscence modules will energy Aurora, the primary supercomputer within the US to anticipate the efficiency of an exaFLOP. The system is predicted to ship in about two years, going past the 2014 Xeon Phi specification.
The US Division of Vitality, Intel and Cray have signed an settlement underneath which the 2 corporations and the Argonne Nationwide Laboratory of DOE will design and construct the Aurora supercomputer, which can enable one-quintillion floating-point calculations per second. The deal has a price of greater than $ 500 million. The system is predicted to ship in 2021.
The Aurora engine shall be based mostly on Intel's Xeon Scalable processors, the corporate's on-board accelerator based mostly on the Xe knowledge middle structure, and next-generation Optane DC persistent persistent storage. The supercomputer will depend on Crays & # 39; Shasta & # 39; structure with Crays Slingshot's connection, which was introduced in supercomputing in November. The system is programmed with Intel's OneAPI and likewise makes use of the Intel-tailored Shasta software program stack.
Two years in the past, the DOE started the Exascale Computing undertaking to drive the event of , software program and functions for ExaFLOP class supercomputers. The corporate awarded analysis grants totaling $ 258 million to 6 know-how corporations together with AMD, Cray, Hewlett Packard Enterprise, IBM, Intel, and NVIDIA. Because it turned out, Intel's method was thought-about probably the most environment friendly for the nation's first exascale supercomputer.
It’s noteworthy that ANL's Aurora supercomputer was based mostly on Intel's Xeon Phi codename Knights Hill in 2014, utilizing the corporate's 10nm course of know-how. The plan was modified in 2017 when Intel turned Knights Hill right into a can in favor of a extra superior structure (and the truth that its Xeon processors have been approaching a Xeon-Phi-like implementation). Apparently, Intel and its companions are assured that the brand new chips will now proceed with the undertaking.
The Aurora supercomputer can deal with each AI and conventional HPC workloads. Presently, the Argonne Nationwide Laboratory says that this machine shall be used for, amongst different issues, most cancers analysis, cosmological simulations, local weather modeling, discovery of medication, and the exploration of varied new supplies.
"There may be super scientific worth for our nation, consisting of collaborations with the Division of Vitality, the Argonne Nationwide Laboratory, the economic companions Intel and Cray, and our shut affiliation with the College of Chicago," stated the director of the Argonne Nationwide Laboratory Paul Kearns. Argonne's Aurona system was developed for next-generation synthetic intelligence and can speed up scientific discovery by combining high-performance computing and synthetic intelligence to resolve real-world issues similar to bettering excessive climate forecasts, accelerating medical therapies, Mapping the human mind and growing new supplies and a greater understanding of the universe – and that's only the start. "