|
Robert Dennard's world is one of limits -- finding them, extending them, surmounting them and, ultimately, trying to ignore them. If the subject turns to his tennis game for example, he waxes poetic about those times when every shot is hit solidly and the ball goes precisely where he wants. "It's like not being constrained by those things we're usually constrained by," he enthuses.
By its very nature, scientific research is aimed at studying the limits of current understanding and taking things a step further. Robert Dennard's invention of one-transistor dynamic RAM (DRAM) was a core development in the launch of today's computer industry, setting the stage for development of increasingly dense and cost-effective memory that continues even today at the heart of every succeeding generation of computers. Among other achievemtns, this was a milestone in a distinguished career that has brought him a host of honors and accolades, including being named an IBM Fellow in 1979, receiving a National Medal of Technology from President Ronald Reagan in 1988, being inducted into the National Inventors Hall of Fame in 1997, and being honored with the Lemelson-MIT Lifetime Achievement Award in 2005.
Dennard, continuing on his path of achievement, subsequently took his work several steps further. Working with co-workers, he developed and verified scaling theory -- an orderly scientific approach to determining and dealing with the challenges posed in designing and building ever-smaller computer devices on silicon chips.
Ironically, Dennard says the inspiration for his development of the single-transistor DRAM emanated from a presentation by in-house competitors in the Research division at a 1966 conference "that really discouraged me" because "they were talking about all the great things they were doing" with thin-film magnetic memory. But their simple approach to a memory element -- "a piece of magnetic material and a couple of lines passing near it" -- made him look for ways to simplify his approach which, at that point, was a complex six-transistor arrangement for storage of a single bit. After working for several months, Dennard reached the realization that a single field-effect transistor and data line could accomplish both the writing and reading of charge stored in a capacitor, and within a year IBM had been granted a patent on one of the key technologies of the computer age.
Robert Dennard was a young Texas-bred scientist (with a Ph.D. from Carnegie Institute of Technology) who thought he knew where he was headed when he arrived at IBM in 1958. He figured he'd "go to IBM and learn a lot of good stuff, and then maybe go out to a smaller company and work my way up. I figured I'd spend about three years." A short chuckle sets up the obvious punch line: "That was 46 years ago."
These days, Dennard could be said to have come full circle on his work, focusing his attention on how far the current silicon chips can be taken, and whether the practical limits of scaling are being approached. Noting that devices now measure as little as 1/10 of a micron (a 100:1 reduction in scale from when he began work in the field more than 30 years ago), he says that "we're getting pretty close to the physical limits, in the sense that as we continue to make them smaller, we don't get the returns in performance and lower power and density because of some fundamental limitations.
"I'm interested in finishing up this thing that I helped get started. And, as we're getting to the limits of this technology, the questions are even more interesting than they've been for a few years, because we're looking for ways around some of the problems. And there are lots of interesting possibilities that need to be evaluated." And limits to be tested.
|
