The speed of the proposed machine, dubbed Blue Gene, is astronomical. It would handle 1 quadrillion operations per second–that’s 15 zeroes–making it the first “petaflop” computer. That’s 500 times faster than today’s scorching supercomputers and 2 million times faster than PCs. Monty Denneau, one of the countless bouncy geniuses at IBM’s labs, told NEWSWEEK that if a conventional PC’s speed were represented on a bar chart as one inch tall, the bar for Blue Gene would reach 30 miles. Added Paul Horn, the head of IBM research, “No one’s dumb enough–or smart enough–to try to do this but us.”
Why would Blue Gene need such speed? The five-year plan is not only an effort to create an entirely new computer architecture. Working with pharmaceutical companies, IBM hopes to solve one of the greatest mysteries of molecular biology: how the proteins that our genes produce fold into their functional shapes. Once scientists better understand that molecular origami, they can simulate the effect of drugs on those proteins and potentially produce cures. The complexity of the task is so confounding that today’s computers can barely simulate a fraction of the folding process. But the lack of computing power, cautions Peter Kim, a biologist at MIT’s Whitehead Institute, is “not by any means the only problem.”
To enter the realm of petafloppery, IBM will combine for the first time several known designs. For starters, the machine will use 1 million processors; today’s big computing iron uses roughly 5,000. But instead of forcing each processor to communicate with the memory system through a relatively slow pipeline, IBM is embedding the memory into the processor to improve the computer’s speed. Second, the processors are small and simple, omitting power-hungry data-storage systems known as caches. And each of the processors will simultaneously handle eight separate series of computer instructions, known as threads. Last, IBM researchers say the communications between chips will reach the kind of speed that could Hoover the Web’s entire contents in a split second.
All this is enough to make researchers “drool,” as one university scientist put it, if Big Blue delivers on its rosy plans. And that’s just the kind of response IBM most wants. Like Deep Blue, the new initiative could produce a windfall of good press while, unlike Deep Blue, having a bigger impact on humanity. IBM also hopes such a splashy project would help lure high-tech talent who have beat a path to Web start-ups.
During the epic chess match, Kasparov uttered that Deep Blue momentarily played “like a god.” If Blue Gene does help solve the molecular mysteries of life, God may be looking over his shoulder.
