The fastest computer in the world today can deliver about 125 petaflops of performance, but that could quadruple in the coming years.
Cray’s XC50, announced Monday, can deliver a petaflop of performance in a single box, and up to 500 petaflops for an entire supercomputer. A supercomputer is typically multiple servers—also called nodes—strung together, combining to provide multiple petaflops of horsepower. So the XC50 needs to be specifically configured to hit 500-petaflop performance, an effort that could take a few years.
There are multiple technologies that boost the performance of the XC50. It is compatible with the Nvidia Tesla P100 GPU and Intel’s Xeon and Xeon Phi processors, which are both accelerators that speed up scientific computing tasks.
A supercomputer in Switzerland called Piz Daint, which uses the older Cray XC30 design, has been upgraded to XC50. It is the world’s eighth-fastest supercomputer, according to a new list of the world’s fastest systems released by Top500 on Monday.
In-system upgrades to Piz Daint are ongoing, and the supercomputer will be merged with another, called Piz Dora. Once the supercomputers are combined and new components in place, Piz Daint will be one of the fastest supercomputers in the world, Cray claims. However, it’s performance numbers weren’t provided by Cray.
Cray’s XC50 is one step ahead in a race to release supercomputers that can deliver an exaflop (a million trillion calculations per second) of performance. The world’s fastest supercomputer is China’s TaihuLight, which can deliver about 125 petaflops of performance, according to Top500.
Piz Daint today has Intel’s older Xeon E5-v3 processors, but will support Intel Xeon E5-v5 processors based on Skylake, coming out in the middle of next year. The new chips will bring tremendous performance increases.
Read more here: www.pcworld.com/article/3141449/hardware/cray-aims-for-the-500-petaflop-mark-with-xc50-supercomputer.html