Japanese supercomputer first to clear 10 petaflops
The fastest got faster, as the K Computer topped its own record on the twice-yearly supercomputer speed test by scoring 10.51 quadrillion calculations per second.
- Shankland covered the tech industry for more than 25 years and was a science writer for five years before that. He has deep expertise in microprocessors, digital photography, computer hardware and software, internet standards, web technology, and more.
The rankings of the 10 fastest machines didn't change at all on today's new version of the Top500 supercomputer list, but the top dog cleared the notable performance hurdle of 10 petaflops.
"Flops" stands for floating-point operations per second and is a measure of how fast a supercomputer can perform mathematical calculations using the Linpack benchmark. The K Computer, at the Riken Advanced Institute for Computational Science in Japan, moved up from 8.16 petaflops, the score it used to reach the top of the twice-yearly supercomputer ranking last June, to 10.51 petaflops.
It reached the new top speed through being fully assembled, with all 705,024 of its Fujitsu Sparc64 processor cores running.
The Tianhe-1A in China, a former No. 1 machine at 2.57 petaflops, remains No. 2. It uses a combination of Intel central processing units and Nvidia graphical processing units. The United States is the top market for supercomputers on the list, but China is in second place, list organizers said.
Intel remains king of the heap when it comes to processor designs, supplying chips for 384 of the systems.
Among other aspects of the November 2011 list:
• Even though the top 10 machines remained the same--a first since the Top500 list began in 1993--there was still plenty of flux elsewhere on the list. The bottom of the barrel for the November list ranked No. 305 in June, and the total performance of all 500 machines rose from 58.7 petaflops to 74.2 petaflops.
• It's now required 50.9 teraflops to get on the list at all. That score beats the top system from 2004. The scores, though somewhat simplify the broad range of computing work that researchers strive to accomplish on these massive systems.
• A total of 39 systems now us graphics chips as accelerators for some computing tasks. That's the approach employed by a forthcoming Cray-built system with AMD 16-core processors and Nvidia Tesla chips slated for shipment to the National Center for Supercomputing Applications.
• IBM supplied 223 of the 500 systems and Hewlett-Packard supplied 146.
• At least 29 of the systems gulp down more than 1 megawatt of power--enough to power 10,000 100-watt light bulbs. The K Computer consumes 12.66 megawatts, but it's "one of the most efficient systems on the list," with a performance of 830 megaflops per watt, the list organizers said.
