The U.S. Government's ASCI program, a driving force behind supercomputing advances for more than five years, has set the goal of achieving petaflop-level supercomputer performance by the year 2010.

ASCI was established primarily to develop systems capable of performing numerical simulation of nuclear weapons detonations, giving the U.S. an alternative to performing actual underground nuclear tests. These simulations work by dividing space into discrete two- or three-dimensional grids and time into small finite intervals or "steps." Compared to today's teraflop-scale supercomputers, a petaflop supercomputer would allow researchers to use denser grids and smaller time steps, dramatically increasing the accuracy and detail of simulations.

Besides nuclear weapons research, the world's fastest supercomputers are also being built with the goal of solving "grand challenge" problems in science. Defeating the world chess champion is one challenge that received much attention in 1997, but this didn't require petaflop performance nor was it an especially critical problem to solve. At least one upcoming challenge, though, is of prime importance to us all.

Computational biology involving human proteins, image courtesy IBM Research media photo collection

The Grand Challenge of Protein Folding

A major focus of the Blue Gene system is the simulation of human proteins and their "folding." Protein folding controls the human body's physical processes, and abnormal folding causes disease. Through simulation of these processes, Blue Gene can help to determine the cause of some human diseases and also contribute to the development of drugs and other means to cure disease.

The Blue Gene/L system, targeted for completion sometime in 2005, will likely be the world's first computer capable of solving the grand challenge of protein folding.

Next page > How Do Supercomputers Perform This Fast? > Page 1, 2, 3, 4, 5