IBM Cheers $325M Supercomputer And OpenPower CPU Win

IBM scored a $325 million win Friday touting not just a massive supercomputer sale to the U.S. government, but also a huge gain for its OpenPower Foundation and development of systems based on the Power CPU architecture.

IBM is set to sell the U.S. Department of Energy next-gen supercomputers that run at blazing 100 and 150 petaflops speeds to power the country's science and engineering research labs as well as boost national security concerning nuclear weapons. The deal is valued at $325 million.

The systems are based on IBM’s Power CPUs and developed in partnership with graphical chip maker Nvidia and high-speed networking firm Mellanox under IBM's nascent OpenPower Foundation established in August 2013.

[Related: HP Launches New Apollo High-Performance Computing Systems]

The systems, named Sierra and Summit, will be used by at DOE at its Lawrence Livermore and Oak Ridge National Laboratories located in California and Tennessee. The supercomputers, according to the DOE, will run seven times faster than previous U.S. government supercomputers.

Both DOE systems have been optimized for "data centric" systems" -- an architecture that embeds compute power everywhere data resides in the system, allowing for a convergence of analytics, modeling, visualization and simulation, driving new insights at incredible speeds," IBM said.

The deal is important to IBM for two reasons. One, it is IBM's highest profile system to come out of its OpenPower initiative that counts Google, Ubuntu, Samsung, Nvidia and Micron as its key members. It's also a boon for IBM's Power8 CPU (12 cores supporting 96 simultaneous threads) which began shipping earlier this year. The use of Big Blue’s Power CPU at DOE is an important proof point in IBM’s ongoing battle against Intel’s x86 chip architecture used in most PCs, servers and high-performance computers today.

’This win cements IBM as a major player in high-performance computing,’ said Chris Pyle, president of Boca Raton, Fla.-based IBM partner Champion Solutions Group. ’It proves if you want to do real computing on a massive scale IBM is still a leader in this space.’

IBM has long maintained its OpenPower Foundation creating a more collaborative chip development environment in contrast to Intel.

’At this point nobody is doing open,’ said Dave Turek, vice president of technical computing for OpenPower at IBM. Other people are making use of commodity technology. With commodity-based technology there is only one supplier. And that's a closed environment. So the opportunity to innovate around that is compressing, not expanding.’

NEXT: Intel Touts Xeon As Highly Customizable

He said IBM is the only OEM building supercomputers in an open design model with the explicit consideration of the role that data plays and simulation and modeling. Turek said that DOE specifically picked its Power systems to replace x86 systems at Oak Ridge lab in Tennessee.

’Some of those previous systems couldn't deal with data the way IBM can,’ he said.

DOE said its Oak Ridge lab will be dedicated to civilian research and its Lawrence Livermore lab would be used for nuclear weapons simulation.

For its part, Intel in September released its 18 core Xeon family of processors (Grantley) emphasizing three workload types, including compute horsepower, storage optimization and the ability to juggle a variety of network workloads. Intel unleashed 32 SKUs and an additional 20 custom SKUs designed for customers and their specific workload needs.

Intel also touts customization as being key to its Xeon processors. The chip is the back-end workhorse for driving compute-intensive tasks such as Internet of Things and handling storage and network-related workloads.

Both IBM and Intel have been touting processors customized for specific workloads.

"In the future, much of the processing will move to where the data resides, whether that’s within a single computer, in a network or out on the cloud," said John Kelly, senior vice president of IBM Research in a blog post. "Microprocessors will still be vitally important, but their work will be divided up," he wrote.

Bringing compute and storage closer together has been a big push within the x86 camp with companies such as Dell touting big gains within some of its converged infrastructure appliances such as one built in partnership with Nutanix.

"Someone can always truncate the distance between storage and compute,’ Turek said. ’Our argument is, IBM will do that on steroids. But we are also going to deploy compute within storage, in the network and all elements of the server hierarchy. If you can get to a point you don't have to move data, you'll be a much happier person. Movement of data creates delay and the delay dominates everything else because there is so much data.’

IBM’s news comes ahead of the IT industry’s SC14 conference for makers of supercomputers. The conference starts Monday in New Orleans.

PUBLISHED NOV. 14, 2014