China’s Tianhe-1A supercomputer just turned in the fastest performance score in the world: 2.507 petaflops. That’s “more than 40% higher” than the score set by the “Jaguar” supercomputer housed here in the U.S. at Oak Ridge National Laboratory, according to the Wall Street Journal.
Part of the reason behind Tianhe-1A’s speed comes from “coupling massively parallel GPUs with multi-core CPUs,” according to U.S.-based graphics chip maker NVIDIA, which supplied 7,168 GPUs (graphics processing units) to the project.
The GPUs were paired with 14,336 CPUs (central processing units) made by Intel. NVIDIA contends that without combining GPUs and CPUs, the speeds attained would have required “more than 50,000 CPUs and twice as much floor space,” and would have consumed 12 megawatts of power. The current system uses just over four megawatts.
The Wall Street Journal notes that since China’s supercomputer uses Intel and NVIDIA parts, “U.S. customers could presumably construct a system with similar performance.” However, other components, such as the communications chips, were designed in China and contain proprietary technology. And China “is also working on its own microprocessors.”
The supercomputer “was designed by the National University of Defense Technology (NUDT) in China” and “will be operated as an open access system to use for large scale scientific operations,” according to NVIDIA.
More on Techland: