IBM TO ANNOUNCE AN ADVANCE IN CHIPS By JOHN MARKOFF c.1998 N.Y. Times News Service IBM plans to announce on Monday that it has perfected a significant new manufacturing process to produce semiconductor chips that can achieve either higher speed or significantly lower power consumption. Known as silicon-on-insulator, or SOI, technology, the new technique is the third major advance in semiconductors that IBM has announced in the last year. The company's top research executives said that the latest advance should give IBM a lead of one to two years in a highly competitive industry. In September, IBM reported that it would begin using copper to connect transistors on its chips, advancing speed. And in June it said it would start making silicon germanium chips, which are expected to have a broad market soon in portable and wireless consumer devices. The rapid-fire string of advances illustrates a major shift in the origins of new technologies and the directions they follow as they are adopted by the computer industry. Until the 1990s, new technologies tended to trickle down from advanced military and high-end corporate applications. But increasingly, the most powerful innovations are deployed almost immediately to the vast market for low-cost consumer electronics. IBM officials said that the realization of SOI technology vindicated a persistent investment in a technology that for three decades had been considered out of reach. ``This is an important strategic advantage for IBM,'' said Richard Doherty, an electrical engineer who worked on efforts to develop the technology at Data General in the 1970s and who is now president of Envisioneering Inc., a computer and consumer electronics consulting firm. IBM engineers have found a way to embed an ultrathin layer of insulating material just under the surface of a silicon wafer. The insulating layer sharply reduces the amount of electronic charge a transistor must store each time it is turned on. That will make it possible for IBM to increase the speed of its chips by as much as 35 percent beginning in the first half of next year. That is equal to the increase in performance typically achieved in an entire generation of chips. The company said it also planned to apply the process in the second half of 1999 to achieve a significant reduction in the power requirements of chips for portable applications. IBM executives estimated that the new process would initially add about 10 percent to the costs of its chip-making process, but they predicted that this could eventually be cut in half. The IBM process uses a machine called an ion implanter to push oxygen molecules at great speeds so they come to rest beneath the surface of a silicon wafer. The secret to the breakthrough, however, turned out to lie in an annealing, or heating, process developed by IBM engineers. By heating the wafers in a particular way after the oxide layer was deposited, the engineers were able to form an almost perfect silicon layer on the surface of the wafer. This silicon layer, only 400 or 500 atoms thick, is the key to transistors that can switch electrical current on and off more quickly or can consume far less energy in battery-powered devices. One significant consequence of the latest advance is that it will help IBM's engineers protect chip circuitry from the troubling effect of natural background radiation, which becomes an increasing problem as the size of transistors shrinks with each new generation of chips. Background radiation can corrupt data stored in memory chips, producing what semiconductor engineers call soft errors. As the chip industry moves toward line widths as fine as 0.18 microns _ about one five-hundredths of the width of a human hair _ soft errors in data are a growing concern. ``A number of major chip makers have simply given up on the technology,'' said Bryan Lewis, a semiconductor analyst at Dataquest, a market research firm based in San Jose, Calif. ``But this will have a dramatic impact on the industry. In 10 years it will have taken over half of the industry.'' Not everyone believes that the technology will give IBM a lead that proves impossible to match, however. An Intel Corp. semiconductor researcher said that Intel had evaluated the technology in the past and decided that it would be a safer strategy to press ahead with pure silicon. Even so, he acknowledged that in low-power applications, IBM's process could provide a real advantage.
The views and opinions expressed in this page are strictly those of the page author.
The contents of this page have not been reviewed or approved by the University of Minnesota.