This Will Be the World's Most Powerful Supercomputer By a Mile

'Aurora' is the next step toward building a computer that matches the power of the human brain.

|
Apr 10 2015, 5:52pm

​Image: Argonne National Laboratory

​The dream of building a computer that can replicate the processing power of the human brain is getting closer. The US Department of Energy announced a commitment of $200 million to build 'Aurora', a supercomputer that could be the next step toward building a machine to match our fleshy, human wits.

Supercomputers capable of processing one billion billion calculations per second—a measurement known as an exaFLOP—are a long-standing goal for organizations building computers capable of mimicking the human brain. The Human Brain Project in Europe, for instance, has set exascale computers as the benchmark for when the human brain can be fully simulated at the cellular level.

No computer on Earth has reached that level of processing yet, but Aurora will get us closer by a leap. When switched on in 2018, Aurora will boast a peak processing power of 180 petaFLOPs; a petaFLOP equals ​one thousand trillion calculations per second. The world's fastest supercomputer as of November 2014 is the Chinese Tianhe-2, which can process 33.2 petaFLOPs. For some perspective, an exaFLOP is three orders of magnitude larger than a petaFLOP.

"This machine—part of the Department of Energy's CORAL initiative—will put the United States one step closer to exascale computing," Department of Energy Under Secretary Franklin Orr said in a statement. Aurora is considered by the department to be a "pre-exascale" machine.

Aurora will be built by Intel and Cray, Inc., the latter of which has set records for computing power with its Titan supercomputer in 2012. Aurora will be based on Cray Inc.'s Shasta supercomputer framework, and will live at the Argonne National Laboratory near Chicago.

Scientists have set anywhere from 2019 to 2023 as the target year for developing an exascale computer, a goal that looked to be in jeopardy when a recent list found thatcomputers weren't getting more powerful as quickly as they used to. A 2018 switch-on date for Aurora could mean that the 2020 target for a brain-level computer isn't so far-fetched.

Although Aurora is being hailed as a step toward computers that can simulate the brain, its main function will be scientific research. Supercomputers are capable of processing huge amounts of data and performing many complex calculations at once, making them perfect tools to solve some of science's toughest problems.

While Aurora will be available to the "entire scientific community," according to the Department of Energy, it says it will be used to develop new materials for more efficient batteries and solar panels, biofuels, and more efficient transportation systems.

Despite a $200 million investment and some of the world's smartest folks on the case, we're still a few years away from a computer that can match the human brain—but perhaps not that many, after all. 

CORRECTION: A previous version of this article stated that a petaFLOP consisted of one thousand million calculations per second, when it's really one thousand trillion. This article has been updated to correct this inaccuracy.