Different is not better At Intel's advanced-chip plants, normal consistency doesn't cut it: The company even copies the air in the room
<http://www.oregonlive.com/business/oregonian/index.ssf?/base/business/1194591355141180.xml&coll=7> Imagine a friend serves you an especially delicious cake and offers to share the recipe. Seems like something you could make, until you get a look at the ingredients: The eggs must come from the same farm where your friend got hers. Flour, ground from the same crop of wheat. Water from the same tap. That's how Intel makes its computer chips: It takes recipes cooked up by its Hillsboro engineers, then copies them exactly at factories in such far-flung locales as Arizona, New Mexico and Israel. Duplicating the technology means replicating the Hillsboro factory right down to the air its technicians breathe. Any change, no matter how slight, could wreck a chip's miniature features. For a powerful new chip that goes on sale Monday, the production strategy that Intel calls Copy Exactly became many times more demanding than ever. Planning to make the chip, code-named Penryn, started a year earlier than usual for a new chip. And the company found Penryn could tolerate even less variation than in previous chip upgrades. "Every day, we're learning something new, and every day you have to be on top of your game," says John Pemberton, manager of a new Intel factory near Phoenix that started making Penryn chips last month. "The level of rigor we need to operate at -- I've been doing this for 25 years -- is really new to me." A race against time Intel, based in California, is the world's largest chip maker and Oregon's largest private employer. For this generation of chips, its Hillsboro engineers created a revolutionary transistor from new materials, making possible chips that are smaller, faster and more efficient than any that came before. Intel says the new technology boosts computing power by a fifth while decreasing power consumption by a third. Computer technology goes stale in a matter of months. So it's a race against time for Intel to duplicate its Oregon success and bring its new chips to the world's computers before new designs from Intel or its rivals deliver more whiz-bang. Intel can't afford to be slow off the mark. And that's where Copy Exactly comes in. By replicating every variable -- from each factory's concrete foundation to the screws that hold production equipment together -- Intel assures itself that each plant will reliably turn out chips from the moment production begins. "We're able to take risks in what we do and what we put in the process, because we know it'll work the same way every time," says Kaizad Mistry, the Hillsboro engineer who managed work on Intel's new chip technology. For the new chip, Intel took some big risks. Circuitry inside Intel's chips had grown so small, just five atoms thick at one key spot, that it could shrink no more without creating breakdowns in performance. Future advances in computing power were on the line. So Intel's Oregon engineers ripped up the conventions of computer chips, replacing the basic material at the heart of the chip with a previously untested cocktail of metals. Their invention delivers improved performance with less energy. Penryn, the first chip using the new technology, will start appearing in high-end computers this fall. By this time next year, most new laptops and PCs using Intel chips will have Penryn inside. Exact everywhere With Penryn's robust computing power and low demands on batteries, Intel hopes the new technology will find its way into new, iPhone-style palmtop computers that are surging in popularity around the world. When Intel announced its new chip in January, The New York Times hailed it on the front page. This month, Time magazine named the chip one of its top inventions of the year. But creating the new chip is just the first step. Although Intel's engineers spent the better part of a decade crafting the new technology, the fevered pace of high-tech innovation will soon overtake the chip. Competitors are far along with their new technologies, and Intel's own demanding two-year upgrade cycle will soon make Penryn yesterday's news. As Penryn moves toward full-scale production, Intel's engineers in Hillsboro have already begun work on successive waves of chip technology, the first of which will crest in less than a year. To make the most of its invention, Intel plans to replicate Penryn millions of times in the coming months at factories across the world. By scattering production, the company taps into its global supply chain and reduces the risk that a natural disaster or other regional crisis could cripple production. Re-creating Oregon conditions elsewhere requires exhaustive attention to detail. Pemberton, manager of the new factory near Phoenix, recalls an unexplained defect that ruined Intel chips at a previous Arizona plant several years ago. Engineers studied their manufacturing tools, their production techniques, the factory layout -- searching for some variation from Hillsboro. Finally, they took a sample of the surrounding air and found high levels of ammonia, emitted by cow manure at a nearby dairy, messing up production. "It literally took a year to solve that," Pemberton says. It's not just the Oregon manufacturing process that Intel duplicates. To a degree, it's also the people in its factories. People such as Daniel Ben-Atar. Ben-Atar, a 42-year-old Israeli engineer, moved with his wife and three sons last spring from their home south of Tel Aviv to Oregon. Along with several hundred of his Israeli colleagues, Ben-Atar spent the summer and much of the fall working inside Intel's huge $3 billion Hillsboro research factory. Cloaked head-to-toe in a white "bunny suit" that helps keep the factory environment pristine, Ben-Atar learned every step of the Penryn manufacturing process from Oregon engineers. When he goes home this fall, Ben-Atar's job is to ensure a perfect match between Israel and Hillsboro chips when Intel starts production next year. A complex challenge Copy Exactly, standard practice at Intel for more than a decade, gives Intel a strategic advantage by saturating the market with Intel's new chips nearly the moment they're ready. "The nice thing about copy exact is you've got the process set in stone," says Jim McGregor, analyst with the research firm In-Stat. "The potential downside is you focus so much on the process that you sometimes overlook the opportunity to focus on the product." Intel played catch-up last year when its much smaller rival, Advanced Micro Devices Inc., beat Intel to market with energy-efficient chips that outperformed Intel's. Although Intel retook the lead and will extend it with Penryn, McGregor says the technology brings new hurdles. Every level of production becomes more complicated. And as Intel enters new markets, such as palmtop computers, it will become progressively harder to predict demand for specialized chips and plan how much of each flavor individual factories should produce. "It is going to be a greater challenge for them, going forward, to balance capacity," McGregor says. Intel's production model is up to that challenge, says Brian Krzanich, Intel vice president for manufacturing and operations. With a solid foundation to build on, Intel can adapt its factories more quickly for specific products, he says, confident in each new step. "Copy Exactly just says: Do what the Oregon guys did, and you know it will work."
