IBM will provide about a thousand servers to Japan's National Institute of Advanced Industrial Science and Technology to build a 11-TFLOP compute cluster.
I remember when 1GFLOP was a big deal. Today, a PPC G4 processor gets that kind of performance when the computer is sleeping. Then 1TFLOP was a big deal.
11TFLOP doesn't seem like a big deal anymore. A couple of years ago, a TFLOP represented years of effort to create a machine built for speed. Doubling or trebling that speed was also a great deal of work. Hardware and clustering has gotten to the point where it's a simple matter of purchasing the hardware and setting up a switched network to get multi-TFLOP performance. We'll see 25TFLOP soon, and all I can say is "that's nice, but who cares?"
