The US Department of Energy (DOE), in collaboration with Cray Inc. and AMD, has announced a new supercomputer, dubbed Frontier, which is touted to be the world’s fastest computer ever.
To be developed at Oak Ridge National Laboratory, Frontier is expected to be available for delivery in 2021.
The supercomputer will be capable of delivering a performance of more than 1.5 exaFLOPS which will exceed a quintillion floating point computations per second. One quintillion is 18 zeros after one. To put things in perspective, the fastest gaming console, Xbox One X, delivers only around 6 Teraflops of performance. The current fastest external GPU Nvidia RTX 2080Ti is also only good for 13.4 Teraflops.
Additionally, the supercomputer is capable of calculating around 50 times faster than current supercomputers. Frontier uses second-generation AI capabilities after the Summit supercomputer launched last year.
Frontier will be based on Cray’s Shasta architecture, and Slingshot interconnects, along with AMD EPYC CPU and AMD Radeon Instinct GPU technology.
The new supercomputer outperforms the recently-introduced Aurora supercomputer which can deliver an exascale performance. Coincidentally, Aurora too has been developed by DOE along with Intel and Cray Inc.
Additionally, Frontier will be able to improve innovations in science and technology and will be used in material studies and also in biomedical and climate research.
Frontier is ORNL’s fourth supercomputer following Jaguar, Titan, and Summit, each machine coming with capabilities that were ‘World’s fastest’ at the time of their launch.
The development of Frontier followed by many supercomputers by the DOE will make the US a pioneer in supercomputers, ready to take on the world with a superfast performance.
Currently, China has the world’s largest number of supercomputers. However, with the arrival of Frontier and Aurora supercomputers, the scale may start to tip in America’s favor.