Of course, a high powered computer needs high power processing, and Nvidia is eager to step up to the plate to create the system. The manufacturer claims GPUs are well suited for this sort of high power parallel computing, and there’s a also new technology on the way that will help make the process even smoother, called NVLink.
NVLink is a new interface protocol for improving the speed and reliability of communication between CPUs and GPUs, which should help diversify the type of processing power that’s included in the supercomputing effort. Nvidia will also leverage the OpenACC toolkit, a system the manufacturer developed that allows software to run on large scale systems without having to rewrite code.
At the moment, the world’s most powerful supercomputer is China’s Tianhe 2, but the second place computer is located in the US Department of Energy’s Oak Ridge National Laboratory. Those machines only produce a theoretical 54.9 and 27.1 petaflops of performance respectively, which is equal to several hundred high-end personal computers working together. An exaflop supercomputer would be capable of 1000 petaflops, almost 20 times the theoretical power of Tianhe 2.
There are currently two major new systems planned for the Department of Energy that are already underway, and those two systems will help pave the way for the larger scale effort. These supercomputers have a number of uses in the big data field, but the upcoming systems specifically are headed for use in the medical field. Between brain mapping, genome, and prescription drug projects, there’s a high demand for more powerful research and development machines.
