Top 10 Supercomputers
10: Nebulae (China)
Nebulae, made by Chinese supercomputer manufacturer Dawning, has been in the top 10 of the TOP500 list since June 2010. The computer has slid from a strong starting rank of No. 2 down to No. 10, but being the tenth most powerful computer in the world is still pretty good, right? With only 120,000 Intel Xeon computer cores, the Nebulae performs at more than 1.27 petaflops and is the second most powerful computer in China.
Nebulae operates at the National Supercomputing Centre in Shenzhen. Part of its speed comes from a number of Nvidia graphics processing units (GPUs) that help accelerate performance. Nvidia’s GPUs also helped China’s Tianhe-1A, currently in fifth place, to best many other super-powerful computers.
Like most supercomputers, Nebulae is tasked with doing what it does best: performing calculations very, very quickly. Its time is reportedly divided between high speed calculations and cloud calculations [source:What’s On Shenzen].
9: Curie Thin Nodes (France)
France’s Curie thin nodes, created by technology company Bull, exemplify how much supercomputer design can vary, even with a lot of the same parts under the hood. The ninth computer on the top 10 list uses Intel’s Xeon processors, but runs on a unique Bull supercomputing suite, instead of the Linux operating system used by most computers on the list.
The computer’s spot on the list is also impressive, given that it uses a mere 80,000 processor cores. That’s 40,000 fewer than the Nebulae! The Curie system peaks at a performance of around two petaflops. According to the French Atomic Energy and Alternative Energies Commission, the Curie computer system is open to scientists through the PRACE research infrastructure, making it one of the most powerful computers in Europe open to research.
8: JuQUEEN (Germany)
JuQUEEN is the first of four computers on the top 10 list running on IBM’s powerful BlueGene/Q platform, with max performance of around 1.6 petaflops. While many computers on the list have been around for the last couple years, the JuQUEEN is brand new. It’s actually a replacement for another system, JUGENE, which was the ninth fastest system on the November 2010 TOP500.
Once JuQUEEN is coupled with another BlueGene/Q supercomputer in October 2012, researchers at the Jülich-Aachen Research Alliance will be able to carve out some time with one of the most powerful supercomputers in the world. Researchers have to submit proposals for projects to justify using some of JuQUEEN’s cores.
7: Fermi (Italy)
IBM’s Fermi is one of five systems the company has ranked on the top 10 list. That’s right — half of the 10 fastest supercomputers in the world come from IBM! Of course, they’re all different, and the IBM Sequoia in the first spot is dramatically faster than many other systems.
Fermi uses more than 163,000 processor cores and, like JuQUEEN, is a brand new European supercomputer set up in June 2012. It’s the fastest supercomputer in Italy and is installed at Cineca, a nonprofit consortium of 54 Italian universities. Fermi maxes out at more than 2 petaflops of performance, making it the second fastest European supercomputer on the list.
6: Jaguar (United States)
The Jaguar is a change of pace. While the last three systems have been newcomers to the TOP500 list, the Jaguar is an old veteran: It’s been on the list since the second half of 2009! That’s practically ancient in the supercomputer world, but the Jaguar, located in Oak Ridge, Tennessee, is a powerhouse. It was the fastest supercomputer in the world on the November 2009 and June 2010 lists. It spent the next year and a half at second and third place before sliding down to sixth.
Jaguar is one of the rare few computers in the top 10 to run on AMD processors, and it also utilizes a special version of Linux, the Cray Linux Environment, built for supercomputers. Supercomputer manufacturer Cray built the Jaguar for $104 million and upgraded it in 2012 to keep it competitive [source:knoxnews]. The company increased Jaguar’s processor count from about 225,000 to nearly 300,000, bringing its processing power from about 2.3 to 2.6 petaflops.
5: Tianhe-1A (China)
China’s electronic pride and joy, Tianhe-1A, was the fastest supercomputer in the world in 2010. It slipped to second place in 2011 and now sits in the number five spot. Like the Nebulae, Tianhe-1A’s performance is bolstered by graphics card maker Nvidia. It uses 168 Nvidia Tesla M2050 GPUs to offload work from its 14,336 Intel Xeon processors. In total, it runs on more than 180,000 computer cores [source: CNET].
Unlike Nebulae, Tianhe-1A is not made by Chinese computer company Dawning. Instead, it comes from China’s National University of Defense Technology. It operates at the National Supercomputing Center at Tianjin. Tianhe-1A can crank out 2.5 petaflops of performance, making it twice as fast as China’s next-best supercomputer.
Despite being another IBM system using Intel’s Xeon processors, the SuperMUC is unique in a couple different ways. Located in Germany’s Leibniz Supercomputing Centre, the SuperMUC uses a new hot-water cooling system to keep the computer’s brain from frying while it’s performing billions upon billions of operations. The SuperMUC is another new entry on the list and performs at up to 3 petaflops, thanks to about 150,000 processing cores.
Efficiency is what really sets the SuperMUC apart: IBM says it’s 40 percent more energy efficient than an air-cooled system would be. They claim the water removes heats 4,000 times more efficiently than air. Thanks to its cutting-edge hardware, the SuperMUC is Germany’s fastest supercomputer. In fact, it’s the fastest supercomputer in Europe, period.
3: Mira (United States)
Here’s where supercomputers get serious. The first seven entries on the list were fast, but not fast enough for top-three ranking. IBM’s new Mira, which becomes fully operational in 2013, peaks at a performance of 8 petaflops. That’s more than twice as fast at the SuperMUC in Germany.
Mira runs on 768,000 processor cores. It’s located at the Argonne National Laboratory, a research laboratory run for the United States Department of Energy. It uses IBM’s BlueGene/Q platform and replaces an older IBM system, Intrepid, which ranked fourth on the list in 2008.
Researchers who submit proposals for the Innovative and Novel Computational Impact on Theory and Experiment program will be able to claim processor time on Mira. Sixty percent of the computer’s capacity will go to their research, while 30 percent will go towards the Advanced Science Computing Research Leadership Computing Challenge. The final 10 percent will be reserved for urgent, time-sensitive computations [source: Information Week].
2: K computer (Japan)
Fujitsu’s K computer, the only supercomputer in Japan that made the top 10, is an incredibly powerful machine. It reigned as the fastest supercomputer in the world on both 2011 lists and now sits at No. 2 with a huge gulf in performance over IBM’s new Mira. It can perform at up to 11 petaflops.
The K Computer is located at Japan’s RIKEN Advanced Institute for Computational Science, where it performs scientific operations including global disaster prevention, meteorology and medical research [source: Fujitsu]. Unlike many of the other supercomputers on the list, it doesn’t run on IBM architecture. The K computer uses Fujitsu’s own SPARC64 VIIIfx octo-core processors. Seven hundred and five thousand computer cores help it churn through operations at an incredible pace.
But believe it or not, the fastest supercomputer in the world is leaps and bounds more powerful than the K computer.
1: IBM Sequoia (United States)
This is the Big Kahuna, the champion on the June 2012 TOP500 list. IBM’s Sequoia is the fastest computer in the world (at least, the fastest visible to the public) thanks to 1.6 million processing cores that can crank out an incredible 16.3 petaflops of performance. Wondering just how incredible that is?
Well, if we look back a mere half decade, to 2008, IBM’s Roadrunner made history (and grabbed the top slot) for cracking 1 petaflop, aka performing 1,000 trillion operations per second [source: IBM]. IBM said Roadrunner was equivalent to 100,000 of 2008’s laptops in performance. And Sequoia is 16 times as fast! Sequoia is one of four computers on the June 2012 list running on the BlueGene/Q IBM design, an 18 core 1.6GHz chip. That’s not an especially fast clock speed by today’s standards, but with 96 racks of chips, the performance really adds up.
What’s Sequoia doing with all that speed, anyway? For a while, IBM has bragging rights — Sequoia is 55 percent faster than the second fastest computer on the list. But they’re putting Sequoia to work, of course. The computer operates at the U.S. Department of Energy’s National Nuclear Security Administration at the Livermore National Laboratory. The computer’s doing important work: One of its responsibilities is simulatingnuclear explosions.