Sony itself is no slouch when it comes to making computers and chips. But what attracts IBM and Toshiba is the PlayStation video-game player, Sony's hottest-selling new product ever. Some pundits point to the launch of Sony's PlayStation2 two years ago as a turning point signaling the end of PC domination of the semiconductor industry. Indeed, the explosive growth of video games is what finally prodded even Microsoft to scrub its old PC-centered view of the world. Last November, it unveiled its own video-game console, called Xbox (see BW Online Special Report, 12/13/01, "The Game Boom").
Blunting Microsoft's deep-pockets competition is one reason for the new IBM-Sony-Toshiba collaboration. To bolster Sony's No. 1 spot in video games, the trio will spend perhaps a half-billion dollars over the next four years to perfect manufacturing methods for a new breed of chips that already are under development by the three cohorts.
BRAINY NETWORK. Thirteen months ago, they teamed up to design what they termed a supercomputer-on-a-chip. Code-named Cell, it will mimic neural cells' ability to interoperate over the brain's network. Cell could turn Sony's PlayStation3, a network-centric upgrade coming around 2004, into a potent multimedia communications system.
PlayStation3 isn't the only target. The IBM-Sony-Toshiba team has decided to go for broke. The latest deal, says Bijan Davari, an IBM Fellow and a technology vice-president at IBM Microelectronics, "is aimed at everything from handheld devices to mainframe computers."
Actually, that decision should come as no surprise. With the size of transistors and circuit lines continuing to shrink, more and more functions are getting stuffed onto each sliver of silicon. By the middle of the decade, many consumer products will consist of only one chip. By decade's end, only large systems like mainframes and aircraft avionics may need more than a single chip.
SYSTEMS ON A CHIP. In this emerging era, the chip will be the product, lacking only software and a housing. That signals "a change to the whole landscape of the chip industry," Davari says.
Today, most silicon engineers design generic chips -- microprocessors, memory chips, graphics accelerators, and display controllers -- that their customers mix and match. To create proprietary products and brand-name value, the makers of computers and electronic systems usually add one or more application-specific integrated circuits (ASICs), which they either design themselves or have designed to their precise specifications.
With system-on-chip technology, all of those disparate skills have to be combined -- and virtually no company can boast of having best-practice expertise in everything from chip design to consumer marketing. For chipmakers, the challenge will be especially daunting. Each chip must be designed from the get-go with input from consumer-savvy engineers and marketers who know the requirements for success in that particular consumer market.
"THEY'RE A MUST." IBM's contributions to the joint effort will focus on the chip technologies developed for its mainframes and servers. These include copper wires and silicon-on-insulator (SOI) fabrication methods. Using copper instead of aluminum for the vanishingly small wires on chips reduces their resistivity, which brings two benefits: Signals zip through the wires faster, and their passage generates less heat. SOI essentially does the same for transistors.
By preventing electrons from leaking into the bulk silicon substrate, transistors switch on and off faster and need less power, meaning less heat. These technologies will be vital for consumer products in the system-on-chip future, Davari insists. "They're a must -- you have to have them."
It's easy to understand why. The world's leading chipmakers soon will start to shrink the width of circuit lines to 90 nanometers, from the 130 nm now used on leading-edge chips. The thinner lines should begin appearing on chips in 2004, according to the technology road map from Sematech, the semiconductor industry consortium in Austin, Tex. Around 2007 or 2008, 65-nm lines should emerge. "As we go from 130 to 90 to 65, the number of transistors will double with each of those cranks," Davari notes.
COOL HAND. Today's chips can get as hot as frying pans, and twice as many transistors might generate enough heat to melt the circuitry. "So we have to reduce the power consumption to keep the heat level flat," he says.
This will be "really difficult," Davari admits. It will take all the power-saving innovations IBM has pioneered -- and then some. Big Blue has a few ideas for new breakthroughs, but it's keeping those cards close to its vest. When the time comes to ante up for the system-on-a-chip poker game, the combined knowhow of IBM, Sony, and Toshiba could be a tough hand to beat. By Otis Port in New York