Intel and the US Department of Energy Commit to Create the First Exascale Supercomputer

Intel and the US Department of Energy Commit to Create the First Exascale Supercomputer

Intel and the US Department of Energy Commit to Create the First Exascale Supercomputer

Intel has announced that they are teaming up with the US Department of Energy to create the world’s first supercomputer with one exaFLOP of compute performance, ushering in the era of Exascale computing. 

At this time the computer is set to be housed in the Department of Energy’s Argonne National Laboratory in Chicago under then Aurora name, utilising Intel technology in the form of future Xeon Scalable processors, Xe graphics cards and DC Optane persistent memory. The PC will be manufactured in collaboration with Cray Inc. 

The computer is designed to handle both HPC workloads and future AI workloads, with the hardware at hand enabling simulations at a level that was impossible before. That said, this system will utilise Intel’s future products, making the system reliant on Intel hitting their performance targets on both the CPU and graphics front. That said, this system will get Intel well into the world of high-performance graphics, with this deal potentially leading the way to future exascale computers if Intel proves successful.    
 

   Intel Corporation and the U.S. Department of Energy will deliver the first supercomputer with a performance of one exaFLOP in the United States. The system being developed at DOE’s Argonne National Laboratory in Chicago, named “Aurora,” will be used to dramatically advance scientific research and discovery. The Aurora supercomputer will be delivered to Argonne National Laboratory by Intel and sub-contractor Cray Inc. in 2021.

 

On the hardware side, Intel has committed to using a future generation of their Xeon Scalable Processors, Xe graphics technology and a future generation of Optane DC memory, all of which will be tied together using Intel’s One API software. The system is set to be delivered in 2021. 

To put an ExaFLOP into perspective, an Nvidia RTX 2080 is said to offer around 10 TeraFLOPS of FP32 compute performance, making an ExaFLOP equivalent to approximately 100,000 RTX 2080 graphics cards. That’s a lot of compute performance.  

You can join the discussion on Intel announcing their plans to create their first Exascale Supercomputer in 2021 on the OC3D Forums.Â