Chip Designers Search for Life
        After Silicon

        By JOHN MARKOFF

           It was a chance meeting between a self-described
           "physicist gone bad" and a chemist. And it may
        someday lead to the creation of a new breed of
        computers based on tiny electronic switches, each no
        thicker than a single molecule.

        Three years ago Phil Kuekes, a Hewlett Packard
        physicist with three decades of experience designing
        computers, was pondering new ways to use a computer
        he had developed using massively parallel architecture
        -- technology that attacks computing problems by
        breaking them into hundreds or even thousands of
        pieces to be simultaneously processed by many
        individual chips.

        At about this same time,
        Kuekes (pronounced
        KEE-kus) happened to
        make the acquaintance
        of James Heath, a
        chemist at the University
        of California at Los
        Angeles whose lab had
        been experimenting with
        tiny structures based on
        molecules of a synthetic
        substance called
        rotazane. It seemed that
        these molecular
        structures might be able to function as digital switches
        -- the basic, binary on-off information gateways of
        modern computing.

        Soon the two scientists were brainstorming about how
        it might be possible to blend Kuekes' computer design
        with Heath's Lilliputian switches. In the fashioning of
        tiny switches, or transistors, from small clusters of
        molecules a single layer deep, the researchers see a
        coming era of computers that would be 100 billion
        times as fast as today's fastest PCs.

        The work, detailed in a paper published Friday by the
        Hewlett-UCLA teams in Science magazine, is a
        noteworthy example of the groundbreaking research that
        suddenly seems to be flourishing at computer
        laboratories around the country -- a flowering of ideas
        that some leaders in the field have begun to consider a
        renaissance in computer science and design.

        In corporate and academic labs, researchers appear to
        be at or near breakthroughs that could vastly raise the
        power and ubiquity of the machines that have insinuated
        themselves into almost every facet of modern society.
        And since many of these efforts, like the
        Hewlett-UCLA research, are focused at the
        microscopic level, computing could become an
        increasingly invisible part of everyday life.

        "A lot of exciting stuff is happening
        outside the mainstream PC market,"
        Marc Snir, a senior research
        manager at IBM Corp., said. "We
        are entering a world where there
        will be billions of small devices
        spread everywhere."

        The Hewlett-UCLA team is just one of six groups
        around the country working under an ambitious program
        of the federal Defense Advanced Research Projects
        Agency that is trying to create a new kind of
        molecular-scale electronics -- known as moletronics --
        that researchers hope will one day surpass the power
        and capacity of the silicon-based technology used in
        today's computers. Last year researchers at Yale and
        Rice universities took earlier steps toward the same
        goal of assembling computers from molecular
        components.

        Meanwhile, separate from the military research
        program, researchers at the Massachusetts Institute of
        Technology's Laboratory of Computer Science are
        trying to meld the digital with the biological by
        "hacking" the common E. coli bacterium so that it
        would be able to function as an electronic circuit --
        though one with the marvelous ability to reproduce
        itself.

        It may all sound esoteric. And even the most
        enthusiastic researchers concede that practical
        applications of their theories and methods may be a
        decade or more away. But researchers seem intent on
        renewing the emphasis on the science in computer
        science, venturing beyond electrical engineering and
        physics to draw upon diverse disciplines like
        biochemistry as they form their hypotheses and test
        them with empirical evidence.

        MIT's computer scientists, for example, are pursuing
        the idea of building -- or possibly growing, if the E.
        coli experiments pan out -- vast numbers of almost
        identical processors that might act as sensors or even
        as the task-performing devices called actuators.

        "We would like to be able to make processors by the
        wheelbarrow-load," said Harold Abelson, an MIT
        computer scientist.

        Abelson and his colleagues, who call their approach
        amorphous computing, are experimenting with the idea
        of mapping circuitry onto biological material. That
        might let living cells function, for example, as digital
        logic circuits. Such circuits are information pathways
        that, whatever their complexity, ultimately involve a
        multitude of quite simple, binary choices: 0 or 1, on or
        off, this but not that.

        Biological cells, of course, would be able to compute
        only as long as they remained alive. But the premise is
        the same as in the molecular-scale work: Pack as many
        millions or billions of these tiny decision switches as
        possible into the smallest spaces conceivable.

        The resulting "smart" materials might be used for new
        types of paints or gels, for example, that would enable
        a highway to be "painted" with computerlike sensors to
        issue traffic reports, or let floors be given a surface of
        infinitesimal actuators that could detect dirt and dust
        and silently and instantly whisk it away.

        In the case of the Hewlett-UCLA work, researchers
        have successfully created digital logic circuits, but not
        yet any in which the molecular switches can be restored
        to their original state -- returning to the off position, for
        example, after having switched to on. And still to be
        developed are the molecular-scale wires that would be
        needed to interconnect the switches.

        The really significant implication of the Science article
        is that for the first time researchers have built
        molecular-scale computing components using chemistry
        rather than the time-honored technology of
        photolithography, the ultraviolet-light etching of
        circuitry onto silicon that is the process for making
        today's chips. The chip industry has not yet reached the
        theoretical limits of photolithography, but the day may
        come when it is no longer possible to etch circuits any
        closer together. That is where molecular chemistry
        could take over -- and possibly wring more computing
        power from a single chip than exists today in the
        biggest, fastest supercomputers.

        Last week, Kuekes said his team had been in contact
        with the MIT group and was now discussing the
        possibility of combining the Hewlett-UCLA molecular
        switching technology with the MIT lab's biological
        processor work with an eye toward future computer
        designs.

        "Think of us as the Sherwin-Williams of the
        Information Age," Kuekes said, referring to the vision
        of suspending billions of tiny processors in a paint.
        "This is the raw material for super-intelligent
        materials."

        The scientists acknowledge that their projects are
        gambles and that any practical applications may be a
        decade or more away. But the work on both coasts
        indicates the breadth of the renaissance now sweeping
        through computer-design circles.

        To some extent, the most recent work is a continuation
        of efforts that began about five years ago and quickly
        grew into the commercial field of
        microelectromechanical systems, or MEMS chips.
        MEMS are microscopic mechanical structures etched
        into the surface of silicon chips; they have spawned a
        multibillion-dollar business, largely around the chips in
        silicon accelerometers that are now standard equipment
        as sensors for air-bag collision systems in cars.

        But as the Hewlett-UCLA and MIT research makes
        clear, even the conventional circuitry of a silicon chip
        -- or the silicon, for that matter -- can no longer be
        assumed. Indeed, computer architects are rethinking the
        whole concept of the microprocessor, the master chip
        that begat the PC and that has been the building block of
        modern computers for a quarter of a century.

        "It's time to do something different. It's time to look at
        the things we've ignored," said David Patterson, a
        computer scientist at the University of California at
        Berkeley.

        In the early 1980s, Patterson helped pioneer one of the
        most significant computer design innovations of its era,
        a technology known as reduced instruction set
        computing, or RISC.

        Building on ideas first advanced during the 1970s by a
        team of researchers working under the computer
        scientist John Cocke at IBM's Thomas J. Watson
        Laboratory, Patterson and his Berkeley graduate
        students proved that sharp increases in the speed of
        processor chips could be achieved by simplifying
        computer hardware and shifting many functions to
        software.

        For almost a decade after the success of the Berkeley
        RISC project, many experts in the computer industry
        believed that RISC would ultimately displace Intel
        Corp.'s X86 computer chip design, on which the
        industry-standard PC was based.

        Hoping to unseat Intel, dozens of new RISC-inspired
        companies sprang up from the mid-'80s through the
        mid-'90s. But beginning with its 486 chip, introduced in
        1989, and continuing in its Pentium series, Intel was
        already incorporating the best RISC ideas into its
        chips, keeping their performance close enough to
        RISC's potential to let the company retain its
        commanding market lead.

        The combination of Intel's hardware and Microsoft's
        software proved invincible, and one by one the RISC
        challengers collapsed. The end of the RISC era almost
        a decade ago left computer designers with few
        fundamentally new ideas for improving computer
        performance.

        But even as the Intel juggernaut moved on, the rising
        importance of the Internet and a growing consensus that
        computing's future lies in inexpensive
        consumer-oriented devices helped fuel the renaissance
        in chip design.

        Patterson, the RISC pioneer, has now embarked on a
        new design approach, known as intelligent RAM, or
        IRAM, that has generated great interest among
        consumer electronics companies.

        RAM stands for random access memory, the
        semiconductor memory that is used as a kind of
        temporary scratch pad by software programs. Patterson
        and his Berkeley colleagues have noted that the largest
        performance constraint in computer systems today is the
        mismatch in speed between the microprocessor and the
        slower memory chips. The Berkeley researchers
        predict that in the next decade processors and memory
        will be merged onto a single chip.

        The IRAM chips would embed computer processors in
        vast seas of memory transistors. Instead of stressing
        pure processor speed, these new chips would place the
        emphasis on avoiding bottlenecks that slow the data
        traffic inside a processor. Such an approach would be
        especially attractive to makers of memory chips --
        companies eager to find new ways to distinguish
        themselves in what is now a commodity market.

        At the same time, some consumer electronics
        companies are pursuing ideas similar to those behind
        the IRAM project. For example, Sony Corp.'s Emotion
        Engine chip, which the company is designing in
        cooperation with Toshiba Corp. for the coming Sony
        Playstation II game console, blends memory and
        processor logic as a way to create faster, more vivid
        on-screen game action.

        But no single computer architecture is optimal for every
        kind of problem. That is why many researchers are
        exploring the idea of reconfigurable chips -- chips
        whose circuitry can be reorganized for each specific
        computing problem.

        One of the most extreme efforts in this direction is a
        processor approach known as RAW, for
        raw-architecture work station, that is being pursued by
        a group of researchers at MIT's Laboratory of
        Computer Science, working separately from the E. coli
        project.

        RAW pushes the idea of RISC to a new extreme. The
        processor would not have an "instruction set" in the
        conventional sense of the term, which refers to the
        types of instructions that let software programmers
        direct a chip to do things like add or subtract, or
        compare and move blocks of data. Instead of using
        preordained instruction sets, RAW would present the
        entire chip to the programmer as a veritable blank slate
        on which tens of millions of transistors might be
        individually programmed.

        Such unprecedented flexibility would mean that the
        same basic chip might be used for an infinite variety of
        purposes -- like creating three-dimensional animated
        graphics, performing supercomputer calculations or
        carrying out tasks not yet conceived.

        "We're trying to determine whether this is simply a
        challenge for programmers or a complicated
        nightmare," said Anant Agarwal, the MIT computer
        designer who heads the RAW project.

        As all these research efforts proceed, computer
        scientists are toiling against the perceived limits of
        Moore's Law -- a guiding principle for chip design
        since the Intel co-founder, Gordon Moore, observed in
        1964 that the number of transistors that could fit onto a
        silicon chip doubles approximately every 18 months.

        But ultimately, there are physical limits to how closely
        together circuits can be etched with light onto the
        surface of a silicon chip and still function -- a reality
        that is expected to mean the end of the Moore's Law
        paradigm sometime around the year 2012. That is why
        forward-looking work like the molecular-scale
        circuitry research of the Hewlett-UCLA team is so
        important.

        "Clearly a technology this radically different won't tip
        over a trillion-dollar industry" like today's
        computer-chip industry, Kuekes said. "But we're
        looking considerably ahead of where silicon runs out of
        steam."