Nuclear mattersWorld's largest supercomputer will be used for nuclear stockpile research

Published 4 February 2009

IBM to build a 20 petaflops supercomputer, called Sequoia, for the Lawrence Livermore lab; a petaflop stands for a quadrillion floating-point operations per second; to put Sequoia’s computing power in perspective, what it can do in one hour would take all 6.7 billion people on Earth with hand calculators 320 years, if they worked together on the calculation for 24 hours per day, 365 days a year

At times, bigger is better. The U.S. government has contracted out IBM to build a massive supercomputer bigger than any supercomputer out there. The supercomputer system, called Sequoia, will be capable of delivering 20 petaflops (1,000 trillion sustained floating-point operations per second) and is being built for the U.S. Department of Energy (DoE). DoE will use the supercomputer in its nuclear stockpile research. The fastest system the department has today is capable of delivering up to one petaflop. The system will be located at the Lawrence Livermore National Laboratory in Livermore, California, and is expected to be up and running in 2012.

The Sequoia system will also be used for a massive power upgrade at Lawrence Livermore, which is increasing the amount of electricity available for all of the lab’s computing systems from 12.5 megawatts to 30 megawatts. This power upgrade will require running additional power lines into the facility. Sequoia alone is expected to use approximately six megawatts.

InformationWeek’s Antone Gonsalves writes that This Sequoia computer is massive indeed (it will cover 3,422 square feet): IBM is building a 500 teraflop system, called Dawn that will help Researchers prepare for the larger 20 petaflop system. The Sequoia system will be using all IBM Power chips and deploy approximately 1.6 million processing cores, running Linux OS. IBM is still developing a 45-nanometer chip for the system that may contain 8, 16, or more cores. The final chip configuration has not been determined yet but the system will have 1.6TB of memory when all completed.

A petaflop stands for a quadrillion floating-point operations per second, and a teraflop is a trillion calculations a second. To put Sequoia’s computing power in perspective, what it can do in one hour would take all 6.7 billion people on Earth with hand calculators 320 years, if they worked together on the calculation for 24 hours per day, 365 days a year, according to IBM. Sequoia is expected to be more powerful than the combined performance of all the systems on the Top500 list, according to IBM.
IBM plans to build this supercomputer at their Rochester, Minnesota, plant. The cost of the system has not been disclosed.