For more than a decade, researchers in labs from Osaka to Yorktown Heights, N.Y., have experimented with a radical technique to double the capacity of computer memory chips. On Sept. 17, chip giant Intel Corp. announced in Japan that it will produce chips using this so-called multilevel-cell technology. Intel isn't actually the first to announce such a revolutionary chip--two startups beat it to the punch. But the $20 billion chip giant's entry legitimizes the new technology.
For now, it's only a quiet revolution, restricted to "flash" memory chips that keep data even when the power is off--and that are used widely in battery-powered products such as digital cameras, cellular phones, and notebook computers. Once the chips find their way into products, in early 1998, you'll be able to shoot twice as many photos with a digital camera--or the same number of shots with double the picture quality. Or you can run larger and more complex programs on a laptop.
BIG STEP. How big a breakthrough are Intel's new StrataFlash chips? Potentially, huge. By using very stable materials and ultrasensitive sensors, Intel's chips can store and read four different levels of voltage inside each transistor, instead of just relying on the binary on-off of conventional chips (table). The net result is that each transistor holds four data values, or the equivalent of two bits, rather than one, in the binary language of electronics.
Where that doubling would really pay off is in the DRAM chips used in desktop computers or in the microprocessors that control them. But such chips use materials that can't easily accommodate multilevel cells and that also require speeds hundreds or even thousands of times faster than flash. Japan's NEC Corp. disclosed in February that it aims to deliver a DRAM chip by 2000 that uses four-level memory cells, but Randall Isaac, research vice-president at IBM's T.J. Watson Research Center in Yorktown Heights, doubts the product's potential. He says most chip researchers think it's unlikely to meet the speed and reliability requirements of commercial memory chips.
Engineers have long dreamed of being able to create microprocessors that deal with more than one bit per transistor. But Intel's flash technology won't help make that happen, either. Microprocessors use an entirely different internal structure, and the limited amount of data they store requires very high-speed circuitry. IBM researchers have toyed with the idea of building microprocessors using multilevel circuits for more than a decade, but, concludes Big Blue's Isaac, "it still isn't viable" for commercial computing.
Still, Intel's new StrataFlash will give it technology bragging rights and a potential edge in flash memory, a market that has grown by 50%, from $2 billion in 1995 to an estimated $3 billion this year. With more and more portable devices needing memory chips, "flash has become a pretty big deal," says analyst Drew Peck of Cowen & Co. Intel now has about one-third of the market--down from 39% at the end of 1995, according to Dataquest Inc. Its new $30 chips, which hold 64 megabits of data each, may help it stage a comeback. As for the revolution, it will have to wait.