Berkeley Lab researchers propose a new breed of supercomputers

Historically, supercomputer makers build larger and more powerful systems by increasing the number of conventional microprocessors — usually the same kinds of microprocessors used to build personal computers. Although feasible for building computers large enough to solve many scientific problems, using this approach to build a system capable of modeling clouds at a 1-km scale would cost about $1 billion. The system also would require 200 megawatts of electricity to operate, enough energy to power a small city of 100,000 residents.

In their paper, “Towards Ultra-High Resolution models of Climate and Weather,” the researchers present a radical alternative that would cost less to build and require less electricity to operate. They conclude that a supercomputer using about 20 million embedded microprocessors would deliver the results and cost $75 million to construct. This “climate computer” would consume less than 4 megawatts of power and achieve a peak performance of 200 petaflops.

“Without such a paradigm shift, power will ultimately limit the scale and performance of future supercomputing systems, and therefore fail to meet the demanding computational needs of important scientific challenges like the climate modeling,” Shalf said.

The researchers arrive at their findings by extrapolating performance data from the Community Atmospheric Model (CAM). CAM, developed at the National Center for Atmospheric Research in Boulder, Colorado, is a series of global atmosphere models commonly used by weather and climate researchers.

The “climate computer” is not merely a concept. Wehner, Oliker and Shalf, along with researchers from UC Berkeley, are working with scientists from Colorado State University to build a prototype system in order to run a new global atmospheric model developed at Colorado State.

“What we have demonstrated is that in the exascale computing regime, it makes more sense to target machine design for specific applications,” Wehner said. “It will be impractical from a cost and power perspective to build general-purpose machines like today’s supercomputers.”

Under the agreement with Tensilica, the team will use Tensilica’s Xtensa LX extensible processor cores as the basic building blocks in a massively parallel system design. Each processor will dissipate a few hundred milliwatts of power, yet deliver billions of floating point operations per second and be programmable using standard programming languages and tools. This equates to an order-of-magnitude improvement in floating point operations per watt, compared to conventional desktop and server processor chips. The small size and low power of these processors allows tight integration at the chip, board and rack level and scaling to millions of processors within a power budget of a few megawatts.

Berkeley Lab is a U.S. Department of Energy national laboratory located in Berkeley, California. It conducts unclassified scientific research and is managed by the University of California. Visit our Website at www.lbl.gov.