News

IBM Computer Achieves Petaflop Performance

A National Nuclear Security Administration (NNSA) supercomputer has achieved an operational rate of 1,000 trillion calculations per second, or 1 petaflop, making the Roadrunner -- which the NNSA commissioned IBM Corp. to build in 2006 for around $130 million -- the world's fastest computer, the agency announced today.

The execution rate is more than double the speed of the top computers on the biannual Top 500 list of the world's fastest computers.

"It's a speed demon," said NNSA Administrator Thomas D'Agostino. "Speed is important here. We have stopped underground nuclear testing" and the new supercomputer will be used primarily to run simulations of nuclear reactions at the subatomic level to help ensure the safety of the nation's nuclear weapons stockpile.

In addition to being fast, the Roadrunner is also energy-efficient. It operates at five times the speed of many other supercomputers while using only one-half the energy, according to Dave Turek, vice president of deep computing at IBM. Before recent developments in architecture, programming and processor design, the energy requirements for operating and cooling such a computer would have been "beyond the pale" for Los Alamos, he said.

The system is housed in an IBM facility in Poughkeepsie, N.Y., in 288 refrigerator-sized IBM BladeCenter racks occupying 6,000 square feet, the company said. It expects to ship the supercomputer to the Los Alamos National Laboratory in July.

The computer will be used primarily to simulate the decay of nuclear materials, but it will also assist with research in astronomy, energy, human genome science and climate change.

It is one of the first computers to deploy large numbers of Cell Broadband Engine processors, designed in a collaborative effort by Sony, Toshiba and IBM for the Sony PlayStation video game console. Turek said the Cell processors used in Roadrunner are a souped-up version of those used in the gaming console, with five to six times the performance but little increase in power consumption.

The supercomputer runs 12,960 Cell processors on IBM BladeCenter QS22 blade servers and 6,948 dual-core Opteron chips from Advanced Micro Devices on IBM LS21 blade servers.

Roadrunner first broke the barrier of sustained petaflop performance on May 25 in a two-hour benchmark calculation. Last weekend, it again achieved sustained petaflop performance while running an Energy Department application, said Michael Anastasio, the Los Alamos lab's director.

"With the exception of two other IBM machines, there are no other machines on the planet that have shown the ability to run a real application at anything over 100 teraflops," or one-tenth of a petaflop, Turek said.

During its first six months at Los Alamos, Roadrunner will be given a performance test with unclassified applications, including simulations of HIV, cellulosic biofuel production, and studies of dark matter and dark energy in the universe. After that, it will be used primarily for classified NNSA work, although some capacity will be made available to outside research and academic institutions.

The NNSA has already ordered an additional one-fourth increase in capacity for classified work.

Featured

  • Sign

    2018 Microsoft Predictions Revisited

    From guessing the fate of Windows 10 S to predicting Microsoft's next big move with Linux, Brien's predictions from a year ago were on the mark more than they weren't.

  • Microsoft Recaps Delivery Optimization Bandwidth Controls for Organizations

    Microsoft expects organizations using its Delivery Optimization peer-to-peer update scheme will optimally see 60 percent to 70 percent improvements in terms of network bandwidth use.

  • Getting a Handle on Hyper-V Virtual NICs

    Hyper-V usually makes it easy to configure virtual network adapters within VMs. That is, until you need to create a VM containing multiple virtual NICs.

  • Microsoft Highlights Emerging Kubernetes Scalability and Governance Efforts

    Microsoft this week highlighted some emerging efforts to improve both the scalability and governance of the open source Kubernetes container orchestration service.

comments powered by Disqus
Most   Popular

Office 365 Watch

Sign up for our newsletter.

Terms and Privacy Policy consent

I agree to this site's Privacy Policy.