How Relevant Are Computing Behemoths?

Reading audio



July 02,2013

A new supercomputer developed by China's National University of Defense Technology has been ranked the fastest in the world by the Top500's semi-annual list of the world's most powerful computers. The supercomputer Titan at the U.S. Department of Energy's Oak Ridge National Laboratory fell to number two. But what are supercomputers? And how important is the race to make them faster and more powerful?

Your basic personal computer has one microchip at its core. The Central Processing Unit, or CPU, executes a set of commands contained in a predesigned program.

The first supercomputers had a few more CPUs, but computer science professor Andrew Grimshaw, at the University of Virginia says that changed as microprocessors became cheaper and faster.

"Today, supercomputers are all what we call parallel machines. Instead of one CPU - central processing unit - they have thousands and thousands," he said. "And in the case of the Chinese machine, depending on how you count, millions of the central processing units.

Parallel machines consist of many individual computers called nodes, situated in one block. They use a lot of power, generate a lot of heat, and require huge cooling systems. Also, supercomputers use programs different from the ones used by ordinary computers.

With enough resources, Grimshaw says, anyone can build a supercomputer to address problems that require millions of mathematical calculations.

But that's not always necessary. A virtual supercomputer can be created by networking individual computers within a university campus or company. These machines then process data during down time, when no one else is using them.

"Those are very easy to run on virtual supercomputers because each problem is independent of all the others and I can scatter these jobs out all around the place," Grimshaw said. "We run these all the time at UVA.

Grimshaw says that until a decade ago, engineers were focused on making computers faster. Since then, he says, they have worked to create more powerful parallel machines.

"It’s transforming science and engineering, and it’s going to continue to transform it in ways I think most people don’t fully grasp - how well we can model and simulate the world now," he said.

Grimshaw says computing ability makes the future of research look, "so bright I gotta wear shades."