US DOE To Employ Intel Xe GPU Tech For First Exascale Supercomputer
It was just announced that Intel and the U.S. Department of Energy (DOE) will deliver the first supercomputer capable of one exaFLOP of compute performance, or a “quintillion” floating point operations per second. The system, named “Aurora”, is being developed at the DOE’s Argonne National Laboratory in Chicago. It is being designed for High-Performance Computing and Artificial Intelligence applications and will be used for a wide array or projects including climate modeling, cosmological simulations, drug response predictions, and a myriad of others.
“Today is an important day not only for the team of technologists and scientists who have come together to build our first exascale computer – but also for all of us who are committed to American innovation and manufacturing,” said Bob Swan, Intel CEO. “The convergence of AI and high-performance computing is an enormous opportunity to address some of the world’s biggest challenges and an important catalyst for economic opportunity.”
The Aurora will use Cray’s next-gen supercomputer system, code-named “Shasta,” and will be comprised of more than 200 cabinets that leverage Cray’s Slingshot high-performance scalable interconnect. The Cray Shasta software stack will be optimized for Intel’s architecture and the system will be based on future Intel Xeon Scalable processors, Intel’s Xe compute architecture, and a future generation of Intel Optane
Intel has been relatively open about its upcoming Xe graphics and compute architecture since its Architecture Day disclosures, but hasn’t revealed any product specifics or mentioned any hard performance targets. We were able to get a number of questions answered in a recent interview with Intel, but it will likely be a few more months before specs are locked down and detailed of the initial products, which are due to arrive in 2020, will be revealed.
According to the release, the contract is valued at more than $500 million and Aurora is expected to be delivered to the Argonne National Laboratory in 2021.