The supercomputer uses 7,168 NVIDIA Tesla M2050 GPUs (graphics processing units) and 14,336 Intel Xeon CPUs and is capable of clocking 2.507 petaflops or 2,507 trillion floating point calculations per second.
The Tianhe-1A will take the top spot from the U.S. Cray XT5 aka ‘Jaguar’ that’s at the Oak Ridge National Laboratory in Tennessee. The Jaguar can clock 1.759 petaflops and is built using 37,376 AMD processors.
The Tianhe-1A is interesting because it combines CPU and GPUs — much like desktop PCs — to create the world’s most powerful machine. In fact, Nvidia, claims if its GPUs weren’t used, then it would have taken 50,000 CPUs and twice as much floor space to create a comparable computer.
The Tianhe-1A was designed by the National University of Defense Technology in China and will be operated as an open access system for large scientific computations.
And from The Guardian:
The next set of rankings is due next week, but Jack Dongarra, the University of Tennessee computer scientist who oversees them, told the New York Times that Tianhe-1 “blows away the existing number one”.
Wu-chun Feng, a supercomputing expert and professor at Virginia Polytechnic Institute and State University, told the NYT: “What is scary about this is that the US dominance in high-performance computing is at risk. One could argue that this hits the foundation of our economic future.”
Professor Arthur Trew, of Edinburgh University, who oversees the UK’s HECToR supercomputer, said the Sino-American battle to have the fastest device was not particularly significant. “They are showing off with big machines – fine. It’s the underlying message that is important. The fact they are pumping this kind of money into building these machines in general is far more important … Europe is being left behind,” he said.