And suddenly, disaster struck! Well, okay, maybe not "disaster," per se, but the latest buzz about "Big Mac," Virginia Tech's supercomputer built from 1,100 Power Mac G5s, is that it might not be nearly the powerhouse that early testing hinted it would be. You may recall that not even a week ago, one of the geeks responsible for ranking the world's top supercomputers revealed that in "early benchmarks," Big Mac was "getting about 80% of the theoretical peak"-- a ratio that, if it held through final testing, might have ranked the cluster second only to Japan's Earth Simulator for raw, unadulterated supercomputing power. The only problem is, now we're hearing that the aforementioned 80% ratio has since rolled over, kicked its little legs up in the air, let out a pathetic little cough, and then burst into flame.

Faithful viewer Karl Kornel forwarded us a New York Times article about the Big Mac project which includes this worrying line: "the Apple-based supercomputer... was able to compute at 7.41 trillion operations a second, a speed surpassed by only three other ultra-fast computers." 7.41 teraflops? Not that that's anything to sneeze at, of course, but Big Mac's theoretical peak is 17.6 teraflops, which means that the 80% ratio has since turned into 42%-- just a slight difference. And before you start thinking that it's just a matter of whom you ask, the Times quotes Jack Dongarra-- the same guy who originally told WIRED about the 80% ratio. 42% is the lowest performance ratio in the current top ten. So what's with the slacking?

The reason why this is so upsetting is less about wasted potential and more about plain ol' bragging rights, pure and simple. Right now the fastest cluster on the charts is a "Lawrence Livermore system consisting of 2304 Intel Xeon processors," which cranks out 7.63 teraflops; sure, it's got 104 more processors than Big Mac and it cost two to three times as much money, but if it beats out Big Mac by even a fraction of a teraflop, the G5 will forfeit a hefty percentage of its inherent nerd points. Despite the fact that their beloved Xeon will still have fallen behind the G5's performance-per-dollar rating by a factor of two or more, Intel groupies will be snickering at us. Insufferable bastards.

It ain't over yet, however; according to the Times, officials at Virginia Tech said that "they were still finalizing their results and that the final speed number might be significantly higher." Here's hoping, because when the official numbers hit the street next month, we really don't want to have to work through the shame of championing a cluster that's dragging its proverbial butt along at a mere 42% of its Jobs-given potential. Quick, somebody hire that thing a motivational speaker!