Faster, more powerful and very expensive. Intel has entered into a $200 million collaboration with Cray to build a new supercomputer. The new computer will be 3 times faster than the world’s fastest supercomputer, being able to process 180 petaflops.
The deal includes two systems: Aurora and Theta.
Aurora is based on Cray architecture. The second system, Theta, is a prototype. While not as powerful as other supercomputers, Theta is still one of the top 10 fastest computers in the world and is able to process 6 PFLOP. Theta is set to be delivered in 2016 and Aurora in 2017 or later.
The biggest issue that both companies currently face is scaling. Under current technologies, it has been historically difficult to maintain a steady growth of speed and processing power. DAEPA states that both companies won’t be able to scale their supercomputers to the 180 petaflop that they would hope to.
Cray and Intel hope to have an exascale system fully completed by 2020 – that is the goal.
The next generation super computers will rely heavily on Intel’s Omni-Path fabric technologies, which allows for unprecedented speeds. Latency issues will need to be corrected and shortened before an exascale system would be beneficial.