HOME / gizmos: The latest gadgets and tech toys.

The Next Little ThingHewlett-Packard's nanochip is so 2015.

By Paul BoutinPosted Friday, Feb. 4, 2005, at 6:26 PM ET

When I walked through the doors of HP Labs, I expected to see a giant 3-D plasma display, a retinal sign-in scanner, and a rolling robot espresso maker. But corporate research and development isn't what it used to be. I get the espresso, but my escort from Hewlett-Packard media relations has to toss a quarter atop the office coffee machine to pay for it. Welcome to Silicon Valley circa 2005.

RSS

The tech sector's penny-pinching ways match up nicely with the cheapo two-guys-in-a-garage vibe that Bill Hewlett and Dave Packard inaugurated in 1939. Senior research fellows at HP's cutting-edge research arm, located just outside downtown Palo Alto, Calif., get cubicles instead of offices. My tour guide makes it clear that the company only funds research it thinks will make HP money, usually in five years or less. No blue-sky projects unless they bring in the black ink.

A lot of the lab's undertakings aim to make money by saving money for customers. My first tour stop is a typical corporate computer room filled with row after row of metal racks. The room is a prototype for a data center that allocates its computers, disk drives, and network on the fly without needing a human operator to constantly reconfigure its machines. Even the air conditioner is wired up—it only blows on computers that are running hot, to keep the electric bill down. Down the hall, other researchers have set up a pretend warehouse complete with a fake loading-dock door. The project's goal is to reduce the cost of misplaced inventory for super-size companies like Wal-Mart. A network of RFID sensors, cameras, and an ultrasonic positioning system tracks packages and shipping pallets better and faster than humans with bar-code scanners.

Upstairs, I find Bernardo Huberman, whose book The Laws of the Web explained the behavior of Web sites and surfers with mathematical formulas. He's now working on using market approaches to increase intra-company efficiency. One of Huberman's goals is to optimize how employees use corporate computing resources like that server room downstairs. The idea: Have employees bid for server time out of their own budgets. The price goes up in the afternoon when everyone's trying to get on, so workers stop running unimportant database queries during peak hours.

Clean room for cool chipsIf you're a former Unix system administrator—like, well, me—then this stuff is super cool. But still, I came here looking for flying robots. They're showing me air conditioners, loading docks, and PowerPoint slides. But in the late afternoon, I'm suddenly jolted awake by a physical gadget straight out of the future—a working nanotech computer chip. One of the few long-term projects at HP Labs is a small nanochip fabrication setup that makes circuits one-third the size of the most advanced chips currently on the market. The nanolab is a clean-room facility, so I have to settle for pressing my nose up against the window. On the other side of the glass, a line of nano-imprinting machines, each the size of a McDonald's fry cooker, represent smaller and smaller scales of development. At the far end of a room about a dozen feet square, a chubby lab worker in an androgynous clean-room bunny suit peers through eyepieces at the latest test run.

The technology is simple to describe, if incredibly hard to produce. Today's microchip plants imprint circuits on coated silicon wafers by shining light through a stencil. Since circuits have now gotten smaller than the wavelength of the laser light that's used to etch them, it's getting very expensive to go smaller using the stencil technique. Instead of using light, HP's nano-imprint machine stamps the circuit design by squeezing a drop of liquid into a mold, then hardening it into a grooved circular disk. After that, it's chip manufacturing as usual—the grooves are filled with platinum to make wire circuits.

Working in miniatureThe group's ponytailed director leads me to a microscope that looks down at one of the finished chips. It's kind of a letdown—I can't see much, other than to verify that the chip is really, really small. Neither the director nor I want to fuss with the microscope and risk breaking the chip, so I just pull my face away and squint down with my two naked eyes. And there it is, lying at the bottom of the microscope tray: a black square with an orderly grid of wires leading into it. It's the smallest computer chip I've ever seen.

There are two roadblocks to making a working microprocessor—or is it a nanoprocessor?—using this kind of technology. First, there's the basic entropy of the universe. No matter how precise the manufacturing process, when you make something this small, about 3 percent of the parts will turn out to be defective. Ninety-seven percent sounds like a pretty good success rate, but for chip makers that's suicidally catastrophic failure, something like having the pilot get on the PA and announce that 3 percent of the parts on your 747 have gone screwy.

Instead of futilely trying to fix those tiny broken parts, HP's scientists have come up with a work-around. One computer architect explained this all to me in terms of "defect theology." It starts with original sin: All components are presumed bad at creation. They must redeem themselves through good works—by passing software diagnostic tests that get run sequentially on groups of components. By testing components in different groups, the system eventually deduces which individual parts are broken. It excommunicates those and carries on computing with the rest.

The second problem with HP's nanochips was that they were too small to include any transistors, the basic component of computer logic circuits. Without transistors, you can have a memory chip but not a CPU, the kind of processor that will boot Windows. But in a paper published in the Journal of Applied Physics this week, three HP scientists explain that they've created logic circuits without transistors using a new building block called a crossbar latch. In short, that means HP's nanochips will perform the exact same functions as today's microchips. It's early to tell, but in 10 years, the tiny chip I could barely see under a microscope might replace the Pentium in your PC. It's smaller. It's smarter. It's sexier. And most important in Silicon Valley these days, it's cheaper.