SILICON DREAMS (int'l edition)Makers of superfast multimedia chips are rushing to map a gadget-filled future that could make the PC obsolete
When Sony's engineers in Tokyo and California were searching for a silicon brain for the company's first video game three years ago, they knew they needed something special to make a splash in the cutthroat $4 billion game-player business. Sony wanted to upstage market leaders Nintendo and Sega Enterprises with scenes so realistic that kids would think they were watching television, not computer-generated images. Sony needed a chip that was both lightning-quick and cheap--because its PlayStation was set to sell for $299.
The price ruled out Intel Corp.'s costly new Pentium chip. Sony Corp. contracted with LSI Logic Corp. in Milpitas, Calif., to whip up a game-player's delight for only $40. The chip's design resembles that used by Silicon Graphics Inc. in its powerful workstations. ``We put more computing power at the fingertips of a 9-year-old kid than NASA used to put a man on the moon,'' boasts LSI Logic Executive Vice-President Brian L. Halla.
Like the Apollo program, this chip and its kin are out to change the world. LSI Logic is one of a howling gang of upstarts determined to strike it rich in a silicon rush that could eventually eclipse the personal-computer market.
The PlayStation has reshaped the video-game market since its debut last fall. Sony has sold more than a million PlayStations in the U.S., and it is now outselling Sega's newest entry, the Saturn, 5 to 1. Sophisticated new chips from the upstart silicon dreamers and from the giants--Sun Microsystems, Hitachi, IBM, Philips, and a dozen others--promise PlayStation scenarios in many more markets, from telecommunications and home entertainment to home appliances.
Newcomers such as MicroUnity Systems, Chromatic Research, and 8X8 believe fortunes await new-era microprocessors that can handle video and sound as readily as today's chips process data. These multimedia chips could also give many common products the uncommon ability to relate to users by means other than a keyboard and mouse.
The galloping success of the graphics-oriented World Wide Web hints at what may happen when microprocessors go multimedia. Take Web browsers such as Netscape Communications Corp.'s Navigator. These programs are based on communications standards that, unlike Microsoft Corp.'s Windows, can work with all flavors of computer chips. Netscape Chairman James H. Clark predicts that by early next year, U.S. Internet access will be built into cable-TV set-top boxes powered by multimedia chips.
According to at least one scenario, fewer and fewer of these chips will be created by Intel, or even wind up in PCs. That opens huge opportunities for companies based outside the U.S. Already, a breathtaking array of digital phones, game machines, ``smart'' pagers, and other communications devices is streaming out of factories in Asia and Europe. Increasingly, the chips inside are designed and manufactured locally. ``We've got this huge momentum gathering,'' says Robin Saxby, managing director of Advanced RISC Machines Ltd. of England, whose chips are already being designed into newfangled $500 Web computers. ``It's no longer a desktop era.''
How the new chips will affect Intel over the long haul is an open issue. But Intel Chief Executive Andrew S. Grove isn't conceding the new markets. In fact, he dismisses set-top boxes, video-game players, and most other smart home products as ``a pimple on PCs.''
The PC, Grove contends, is destined to become the ubiquitous information appliance, the hub of the digital universe--including homes without Windows. Intel's new Pentium Pro packs more than enough oomph to function as a combination game machine, set-top box, and multimedia phone, plus some other still-nascent products (page 57). ``They'll all be subsumed by the personal computer,'' Grove insists.
So the lines are drawn: It's the PC vs. virtually every other gadget that processes digital signals. Although Intel's core business isn't at stake, some experts doubt the company can make good on its grander plans for the PC. History shows that new technology always expands rather than limits options, says Forrester Research Inc. analyst William Bluestein. ``So it won't be all in one machine.'' Besides, it's unlikely one company will continue to so dominate an industry that could top $300 billion in the year 2000. Says venture capitalist William Davidow, a former Intel marketing chief: ``There are going to be a lot of crumbs--and they're going to be $100 million to $1 billion crumbs.''
No matter the outcome, this much is certain: Consumers are in for a treat. The new microprocessors ``are going to spawn new applications and new products that you've never dreamed of,'' says Mark G. Stephenson II, a marketing vice-president at Philips Consumer Electronics. Many will emerge in a huge market that remains almost virgin territory for microprocessors, traditional consumer electronics. In the U.S., this is a $60 billion business, making it larger than the $48 billion PC business.
MICROBRAIN BOOM. For example, despite a slow start for interactive TV, market researcher In-Stat Inc. sees the market for interactive digital set-top boxes soaring to 13.4 million units in 1999, up from only 600,000 boxes in 1994--an average growth rate of 86% a year. The surge in video-game players will be even more spectacular. Market watcher Dataquest Inc. predicts that sales will rise from 4.8 million units in 1995 to 17.9 million units in 2000. Dataquest also predicts sales of handheld computers and personal digital assistants will soar almost sixfold from 1994's total, to 5.6 million units in 1999.
All of which will send microprocessor sales zooming. By 1998, these chips will account for 27% of total semiconductor sales, pegged at $242 billion by In-Stat. Since 1992, they've hovered at around 23% of total sales. After 2000, microbrains seem sure to outgrow all other kinds of chips. The reason is a new phase of ``silicon integration.''
In the 1980s, electronic systems were built around a control chip plus hundreds of companion chips. Today, the brain is still there, but the number of companions has shrunk to a couple of dozen. And because transistors are so much smaller, chips can be packed with 10 to 20 times as many of them as a decade ago. As transistors continue to shrink, products will soon have little else but a microprocessor, sometimes throwing in a memory chip or two. So, almost all chipmakers are scrambling to get in on the action in microprocessors. Without a brain chip, the future looks bleak.
LSI Logic has invested heavily in recent years, getting ready for one-chip systems. Founded in 1980 by Wilfred J. ``Wilf'' Corrigan, a British immigrant who worked his way up to chief executive of the pioneering Fairchild Semiconductor Corp., LSI Logic has assembled a design library of various microprocessor ``cores'' plus dozens of other modules. These can be mixed-and-matched to produce a different chip tailored for each specific job.
Probably the most radical restart comes from MicroUnity Systems Engineering Inc. in Sunnyvale, Calif. The company has spent the past seven years reinventing the microprocessor. This year, it plans to market a supercomputer-on-a-chip with a fundamentally new ``architecture'' that can crunch multimedia signals with dizzying speed. This ``mediaprocessor'' will use parallel processing on a single chip to whiz through streams of video, sound, and data at 1,000 times the rate of today's brain chips. The future belongs to mediaprocessors, says MicroUnity Chairman John Moussouris, because the Web's soaring popularity proves the age of computing is giving way to a new age of communications. ``The PC architecture is nearing the end of its life cycle,'' he asserts.
``CRAY ON A TRAY.'' If MicroUnity delivers on its promises, its chip will be not only blazingly fast but also inexpensive enough to be used in budget cellular phones. Moreover, upgrading the chips will be cheaper still: Just load the chip with smarter software. That means users of systems built around a mediaprocessor won't be trapped on a PC-style treadmill, where new hardware is required every few years to run the latest software.
Older chipmakers see the same opportunities. Take the TriMedia chip unwrapped last year by Philips Semiconductors, the chipmaking unit of Dutch giant Philips Electronics. It can process multimedia signals at an incredible 4 billion operations a second. That's enough to choke 10 Pentiums. Texas Instruments Inc. has designed a similarly swift chip. Advanced RISC Machines, a six-year-old British company, is focusing on low cost instead of the highest performance with a sub-$20 chip that runs Apple's Newton, 3DO's game player, and Oracle's recently announced $500 Network Computer.
Such lickety-split speed and nimbleness reflect the chips' ability to skirt the near-monopoly of ``Wintel'' PCs--those with Intel chips running Microsoft Windows. Since Microsoft compatibility isn't required, the chips don't need the so-called complex instruction-set computing (CISC) technology pioneered by Intel 25 years ago. Instead, they use more efficient methods, such as reduced instruction-set computing (RISC).
Upcoming digital videodisk (DVD) systems will all rely on RISC chips, and they promise to spark a buying spree among VCR owners who want the superior image quality of today's videodisk players plus the ability to record video signals. Thomson Consumer Electronics figures millions of DVD systems will be snapped up over the next three years, and its second-generation models may mate DVD with digital satellite service and software for surfing the Internet.
To make its TVs smarter, Philips plans to use the RISC design from Silicon Graphics' MIPS Technologies unit. Philips' digital TVs will enable sports fans to summon their own instant replays from images stored on memory chips. And cable TV's next-generation set-top boxes will link home screens to online data banks brimming with movies, travel guides, and assorted digital paraphernalia. Tapping these services in a way that won't intimidate consumers will require extremely smart chips. In fact, tomorrow's set-top box will be stuffed with so many transistors that John C. Malone of cable giant Tele-Communications Inc. has dubbed it ``a Cray (supercomputer) on a tray.''
Yet the chips will be so cheap that even ordinary objects like shoes will use them. At Massachusetts Institute of Technology's Media Laboratory, 40 blue-chip companies, including AT&T, Walt Disney, and Nike, are pumping $5 million annually into a project dubbed Things That Think. The goal is to ``educate'' dumb products for an Internet-worked world where heels have chips that monitor the news in signals flowing through carpets--and alarm clocks will know, from checking your digital calendar, to wake you an hour earlier when you have a dawn flight to catch.
Can the PC somehow subsume, as Grove maintains, such emerging technologies as Sun Microsystems' Java chips and software? Java chips are designed to give pocket-size digital pads and even cellular phones the ability to run remote Java-software programs stored anywhere on the Internet. MicroUnity's Moussouris doesn't think the PC can win it all. ``The communications world is going to change a lot of the rules of the game,'' he says.
What's for sure is that future systems will deal with vision and sound just as adroitly as Intel's fastest Pentium handles data. And that's going to yield a bumper crop of smart products. VCRs and PCs with infuriating controls will be remembered as clumsy relics.
By Robert D. Hof in San Francisco, with Otis Port in New York
Updated June 14, 1997 by bwwebmaster
Copyright 1996, Bloomberg L.P.