Gary Taubes
In some areas, the payoff of a petaflops computer--more than 1000 times faster than the fastest existing machine--will be surprisingly modest: only a sixfold increase in spatial and temporal resolution in climate models and other three-dimensional simulations. But a petaflops also marks a threshold where other kinds of computation will become feasible or practical for the first time. They include time-critical problems such as computational support for surgery and applications combining many different kinds of simulations in a computational world, in which multiple users could run simultaneous experiments.