But the big news is IBM's prototype Blue Gene/L, being developed for the Livermore Labs. It currently runs at 70.72 teraflops --but they're still tweaking it. When the final version is finished, it should hit 360 teraflops without breaking a sweat.
The current record? Two years ago, 35.86 teraflops. The newest computer, IBM's Blue Gene, potentially could increase the record by a factor of 10. What are they going to use this for, you might ask. Well, at least they're doing the right thing. Blue Gene would (finally) be powerful enough to do protein folding, something which is currently being done through distributed computing (and a very worthwhile cause, I might add). Protein folding could lead to huge advances in the understands and eventual treatment of Alzheimer's, Mad Cow (BSE), CJD, ALS, Huntington's, and Parkinson's disease.
Computer science (the real science part) is finally producing machines that can do things not just that humans cannot do, but doing things we never thought we'd desire to do. Sure, playing chess against a machine is a novel activity, but divining intelligence from such an encounter is another thing altogether. When we do reach the computing power of a human brain, what then? It's not like we'll stop development of computers; we'll obviously surpass that level quickly. And that's a big deal. For the first time in history, mankind will have access to answers from an intelligence more powerful than himself (faith-based Q&A aside).
How does that molecule act when we drop in, say, this new chemical? Does the reaction differ in someone who is taking drug A? The computer will tell you, accurately and relatively quickly. Whole buildings will be devoted to simulating individuals many times over, simply to test reactions to, well, anything we can imagine.
It's a very exciting, and scary, time.