Virginia Tech researcher Wu Feng hopes his work on the HokieSpeed supercomputer will help make supercomputing more accessible (Photo:Virginia Tech)

When Wu Feng looks at an iPad, he sees something more than a great way to play Fruit Ninja. To him, Apple’s sleek device looks more like a compute node on a supercomputer of the future: 1.5 gigaflops of computer power just waiting to be harnessed.

Feng — an associate professor of computer science at Virginia Tech University — hopes to one day bring the supercomputer to an entirely new audience. Today, supercomputers are used by research laboratories, oil companies, and big financial firms, but Feng wants to put them in small businesses and doctor’s offices. And he plans to do it using existing consumer hardware.

He’s known for building very small supercomputers. Nine years ago, before anyone cared about power consumption, he helped build a 240-node supercomputer that used as much power as two hair dryers. Now, he and his fellow researchers are assembling the software and know-how that will help make supercomputing technology something that just about anyone can use. Their test bed is a 209-node computer called HokieSpeed.

By today’s standards, HokieSpeed is not exactly an elite supercomputer. It cost just $1.4 million to build, and last month, it was ranked 96 on a list of the world’s fastest supercomputers.

But Feng and his fellow researchers developing techniques and software to allow HokieSpeed take maximum advantage of its 2,500 Xeon central processing units and its 185,000 Nvidia graphics chips. These are the same chips that ship with any desktop computer, and Feng thinks that if his team gets the development tools right, desktops could soon be used as supercomputers. “You can imagine that HokieSpeed in the next seven years will literally be inside someone’s personal computer,” he says.

Once built only with the kind of high-end computer processors seen in expensive servers, supercomputers are now commonly built with the same type of Intel processors that you find on a desktop computer. And in recent years, scientists have figured out new ways of using the graphics chips you find on desktops, too. It turns out that these graphics processors are remarkably good at certain types of mathematical operations.

Think of it this way: in 1993, the most powerful computer on the planet was a 60 gigaflop Thinking Machines supercomputer operated by Los Alamos National Laboratory. A gigaflop is one billion calculations per second. At 1.5 gigaflops, a single iPad 2 would have ranked around number 177 on that list.

“A lot of the capability, right now, of personal computers is just latent,” Feng says. “It’s just waiting to be exposed and extracted.”

Feng likens the Nvidia graphics processing units (GPUs) to race cars. They burn through fields of data, performing simple instructions at a blazingly fast rate. The Xeon central processing units are more like an SUV. They’re slower, but they can follow really complicated instructions, venturing into territory that would gum up the GPUs.

The Virginia Tech team is figuring out how to best to farm out computing jobs so that the GPUs and CPUs do what they do best, without ever going idle, and without spending too much time communicating with one another.

It’s not easy, but they’re using HokieSpeed to build tools for designing and compiling software so that it can be tweaked to run fast on these systems. They’ve also built what they call an “automated runtime system,” which works with the supercomputer’s operating system to speed things up even further.

What might a small-business supercomputer look like? It might be a cheap desktop system that does complex design and simulation work, so that companies don’t have to build as many real-world prototypes of their new products. Or it might be a number-crunching medical device that helps a doctor deliver lab results while the patient is still being examined.

Just as the Internet and cheap servers gave small companies a way to compete on the web, Feng hopes that the Virginia Tech technology can help out anywhere there’s number-crunching. “I think computing, likewise, is something that if small businesses can leverage in the appropriate way, they can level the playing field against their bigger brethren.”