Intel and Cray team up to build a $500m exascale supercomputer by 2021

The two companies are developing the Aurora program, with a goal of putting together an exascale computing system for U.S. DOE’s Argonne National Laboratory

Intel and Cray have recently announced that the two companies are developing the Aurora program, with a goal of putting together an exascale computing system by 2021 for the Argonne National Laboratory. Apparently, the prefix ‘exa’ indicates bigness, which equates to 1 quintillion floating point operations, or FLOPs, in this case.

An average modern CPU, in comparison, does about hundred or more gigaflops. A teraflop is made up of a thousand gigaflops, a petaflop is made up of thousand teraflops, and an exaflop is made up of a thousand petaflops. Therefore, despite major enhancements in computing efficiency utilized for making super powerful desktops and smartphones, the supercomputer is making a difference of several orders of magnitude.

Supposedly, this kind of computing power actually has use cases in today’s world. One of the recent instance of computing limitations in the real-world research was a study about how could climate change affect cloud formation in specific regions, to support the trend and lead to a vicious cycle.

Sources familiar with the matter cited that estimations for these kind of thing could only be done with lot coarser models earlier and computing resources were quite tight for allowing the extremely large number of variables that are involved here.

By having more computing resources, the more could be dedicated towards discovering materials for creating organic solar cells with higher efficiency, discovering new approaches for prediction of drug response, and building extreme-scale cosmological simulations, as Intel stated in a press release.

Intel mentioned that Aurora would be the first exaflop system in the United States, a significant caveat, as China aims to successfully complete the task a year earlier. Also, there is no reason to think that China would not achieve it, as the supercomputers from China have reliably been one of the fastest in the world.