backtop


Print 26 comment(s) - last by DeepBlue1975.. on Jun 24 at 9:23 PM


HPC system at the Center for Nanoscale Materials, Argonne National Laboratory  (Source: Argonne National Laboratory)
Intel might have the lead in sales, but the most powerful supercomputer in the world uses IBM chips with AMD Opterons

Every year since 1993, the TOP500 project has kept track of the 500 most powerful supercomputers in the world. It started off as a way to chart the growth of High Performance Computing (HPC), but has now evolved into a technological and marketing battleground. The bragging rights for the fastest and most powerful computer in the world have been contested by various countries, manufacturers, and architectures.

The TOP500 list itself is released by the project twice a year, and is compiled primarily by Hans Meuer of the University of Mannheim in Germany, and Jack Dongarra of the University of Tennessee. Located at top500.org, the list is sortable by vendor, country, geographic region, and much more. The latest update has just been announced today at the opening session of the International Supercomputer Conference in Hamburg, Germany.

The most powerful computer in the world is still IBM's Roadrunner, located at the US Department of Energy's Los Alamos National Laboratory in New Mexico. Built at a cost of $133 million, it was the first system to break the petaflop barrier (1 quadrillion floating point operations per second, or 10^15) on the TOP500 project's Linpack benchmark.

Roadrunner is a unique hybrid design, using 12,960 IBM PowerXCell 8i CPUs combined with 6,480 AMD Opteron dual-core processors. It uses Red Hat Enterprise Linux along with Fedora as its operating systems. Roadrunner is being used primarily to simulate how nuclear materials age in order to ensure the safety of the US's aging nuclear arsenal.

In second place is Cray's XT5 Jaguar system installed at the DOE’s Oak Ridge National Laboratory. Jaguar reached 1.059 petaflop/s shortly after its installation in 2008, but no further measurements have been possible since then due to its heavy workload .

A new IBM BlueGene/P system named JUGENE, recently installed at the Forschungszentrum Juelich (FZJ) in Germany has claimed third place. It achieved 825.5 teraflop/s (trillion floating point operations per second) on the Linpack benchmark, but has a theoretical peak performance of just above 1 petaflop/s. FZJ is also home to the new tenth place system named JUROPA.

These two systems are the only non-US based HPC systems in the top ten, with the rest of the top ten slots being occupied by American supercomputers located at American universities and research centers. Not surprisingly, 291 of the TOP500 systems are located in the United States.

Designs vary widely; 383 of the TOP500 systems use quad core processors, and 4 systems use the same Cell 9-core CPU found in Sony's PlayStation 3 video game console. Two systems built by Cray are already using new six-core Istanbul AMD Opteron processors. Almost 80% of the TOP500 systems use Intel processors, and several systems are already being planned using Intel's eight-core Nehalem-EX Xeon server CPUs.

HPC systems are typically used in large corporations, universities, and research institutions. They are often used for basic and applied scientific research, such as modeling the decay of nuclear waste, simulating tectonic stresses to predict earthquakes, examining protein folding, and weather/climate modeling. Corporations are increasingly using cluster-based supercomputers for applications such as virtual prototyping/modeling, data mining & analysis, transaction processing, and video rendering.

The future for HPC development looks bright, as IBM is planning its Sequoia system for deployment in 2011. It will be capable of 20 petaflop/s, making it faster than all of today's TOP500 systems combined. Although that seems powerful, a 1 zettaflop/s (one sextillion FLOPS, or 10^21) system would be needed for accurate global weather modeling.

The next update to the TOP500 list will be presented in November at the IEEE Supercomputer Conference.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: One Day
By eddieroolz on 6/24/2009 2:29:21 AM , Rating: 2
Along with using nearly 20,000 processors to accomplish that.

Let's hope in 10 years time, we can drop in a processor and get 1 petaFLOP at only 95W.


RE: One Day
By ikkeman2 on 6/24/2009 7:31:45 AM , Rating: 3
95W - That won't run in my HTPC case


"We are going to continue to work with them to make sure they understand the reality of the Internet.  A lot of these people don't have Ph.Ds, and they don't have a degree in computer science." -- RIM co-CEO Michael Lazaridis














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki