June 17, 2008 A collaboration between IBM and the Los Alamos National Laboratory has resulted in the world's fastest supercomputer. Roadrunner can run at speeds above the "petaflop barrier" of 1,000 trillion operations per second, making it twice as fast as IBM's Blue Gene/L™ and opening up an era of science at a previously unseen scale.
Roadrunner began operating earlier this month at Los Alamos National Laboratory in New Mexico where around 80% of its capacity will be used to ensure the safety and reliability of the US nuclear weapons stockpile. In addition, the supercomputer will also be used for research into astronomy, human genome science and climate change, with researchers already using the machine to mimic complex neurological processes of the human brain that were previously beyond the reach of artificial replication.
NEW ATLAS NEEDS YOUR SUPPORT
Upgrade to a Plus subscription today, and read the site without ads.
It's just US$19 a year.UPGRADE NOW
The machine uses "hybrid" computing architecture, combining conventional AMD microprocessors (like those found in consumer laptops) with cell broadband engines (like those used in the Sony PS3), but on a massive scale. In all, Roadrunner connects 6,948 dual-core AMD Opteron™ chips and 12,960 PowerXCell 8i processors, with the Opteron processors used for standard processing and the PowerXCell 8i engines accelerating mathematical and CPU-intensive processing.
The numbers are mind-boggling - occupying 6,000 square feet, the Roadrunner has 10,000 connections require that 57 miles of fiber optic cable and its 80 terabytes of memory weighs 500,000 pounds.
Apart from the incredible scaleability of the system, the hybrid computing approach also delivers a relatively affordable and energy efficient system, one which IBM believes will find its way into general commercial use where, for example, it could be used to simulate the inner workings of an entire aircraft in one setting, model the effect of pharmaceutical drugs on the human body or applied to extremely complex financial modeling.