February 2005

Intel's Centrino Solution

The old mantra for PC chip makers -- faster is better -- is breaking down. Can platformization take its place?

By Wade Roush

Before 1991, only a few computer aficionados cared which company made the micro­processors inside their PCs, or how fast those processors ran. But then came "Intel Inside," the chip maker's ingenious campaign to mar­ket directly to consumers. The adver­tising crusade not only trained PC buyers to look for the Intel sticker on new desktops and laptops; it made them feel old-fashioned if they didn't have the latest, fastest edition in the 486 or Pentium ­series of chips. And Intel prospered, cementing its lead over rivals such as Advanced Micro Devices. An impressive 82 percent of PCs shipped globally in the third quarter of 2004 contained Intel ­microprocessors.

But computing is changing in ways that are forcing the company to stretch beyond its traditional talent for making and marketing faster and faster microprocessors. For one thing, there's a ceiling on the number of transistors that can work side by side on a single chip without overheating, and Intel and its competitors are already banging up against it. That's leading to more efficient designs that use multiple processors and get tasks done faster by breaking them up, rather than by making each processor do more operations per second. At the same time, people are taking advantage of innovations like wireless broadband to use their computers in new ways. If your laptop's main function is to keep you in touch with the office from any airport lounge or conference hall, you'd probably prefer an energy-efficient processor that gives you an extra hour of battery life to one that can run your Power­Point animations faster.

Intel engineers saw these trends coming, and shortly after the millennium they started designing a cooler-running, more power-efficient processor called the Pentium M. When wireless Internet standards began to catch on with consumers around 2002, Intel decided to combine the Pentium M, a Wi-Fi radio, and a new low-power chipset -- a group of memory and graphics chips supporting the CPU -- into a package called Centrino. Since the launch of Centrino in March 2003, Intel has captured a respectable 11 percent of the market for wireless networking (see "A Newcomer to Wireless," below), up from zero prior to Centrino.

Centrino can be viewed as a case of good timing (Wi-Fi gained popularity just as Intel was looking for ways to make laptops and notebooks more useful) or as a brilliant encore to "Intel Inside." But executives say the project's real significance lies elsewhere. With Centrino, the company set itself a threefold challenge: to build on the strengths of formerly separate Intel products such as microprocessors and wireless networking cards by tailoring them to work in unison; to coördinate the work of the divisions responsible for these components, so they could be launched simultaneously under a single Intel brand; and to convince PC manufacturers and consumers that they still need Intel technology in a market where mobility and communications, rather than simple computing speed, are paramount.

Incoming CEO Paul Otellini and others -- who apologize for not coming up with a more elegant term -- call their new philosophy "platformization." They say sequels to Centrino are coming in areas such as home digital entertainment and enterprise computing. Platformization "means the convergence of computing and communications," says chief tech­nology officer Patrick Gelsinger. "But it's far more than that, because it changes every aspect of what we do."

The first signs of change came early in this decade, when designers at Intel admitted to themselves that "eventually the microprocessor would hit a power wall," says Mooly Eden, then vice president of Intel's Mobile Platforms Group and the man most credited with conceiving and executing the Centrino project. Making computers more portable meant reducing the CPU's power consumption and heat emissions, which in turn meant sacrificing some clock speed. The first step in that direction was the Pentium M, which used less power and generated less waste heat than its predecessor, the Pentium III, but ran at only 65 to 85 percent of the Pentium III's top clock speed.

The second step: the so-called 855 chipset surrounding the Pentium M, which also used less power and was small enough to squeeze into a notebook-sized computer. As a pair, the Pentium M and the 855 chipset would have allowed computer manufacturers to sell laptops that stayed on for more than an hour longer, without any improvements in battery capacity (which remains the biggest laggard in mobile computing technology).

But then came a third development. As Eden puts it, "Now that you can go away from your desk for five hours, you must make sure you are still connected." Wi-Fi was an almost ideal way to do that: it provided communication at DSL speeds and allowed users to connect to the Internet from anywhere within a 100-meter radius of a central base station (or 400 meters outdoors). At the time, connecting to a Wi-Fi network meant buying a separate, removable network card from a company such as Broadcom. But Eden and his colleagues realized that if they could build Wi-Fi chips small enough to fit inside a laptop, Intel would have all of the components needed to make laptops into mobile workplaces. So even though the three com­ponents of Centrino "were not conceived as a 'platform' from day one," says Eden, the idea of cobranding them took hold quickly. "Paul Otellini was very strong about going in this direction, and there was no argument between Paul, Craig, and Andy," Eden says, referring to Craig Barrett, the company's sitting CEO, and Andy Grove, its cofounder and chairman.