World's Most Powerful Supercomputer Gets Even More Powerful, Breaks 10 Petaflop Barrier
How do you make the most powerful supercomputer in the world even faster? It’s simple, really. Just keep adding components! When Japan’s K supercomputer assumed the top slot back in June, it did so thanks to a team of 68,544 CPUs working in tandem to achieve a maximum LINPACK performance rating of 8.162 petaflops. Since then, the last of its 864 racks have been installed – and that extra firepower has boosted K’s performance over the 10 petaflop barrier.
All in all, the final configuration ended up with a total of 88,128 CPUs, which managed to achieve a measured LINPACK performance of 10.51 petaflops in just around 29.5 hours, with an execution efficiency percentage of 93.2 percent (a 0.2 percent improvement over the June benchmark scores). In layman’s terms, K kicks ass, and it kicks ass harder than any other computer on the planet.
Fujitsu, who developed the system with Japan’s RIKEN institute, bragged in its press release that K isn’t just brawn; it’s also remarkably resilient for a supercomputer. “It also demonstrated an extraordinary level of stability for one of the world's largest-scale systems, as the overall system comprised of 88,128 CPUs ran without a single failure for 29 hours and 28 minutes.”
RIKEN and Fujistu say they are still tweaking the operating system, compiler, and other system software for even better performance in the future. They hope to have it ready for actual research use (and not just benchmark-shattering LINPACK tests) by November of next year.
Update: In the original publication, Toshiba was incorrectly identified as having developed the system alongside RIKEN. The article was altered to give proper credit to Fujitsu.