Ruiz has made sure AMD can adjust the type of chips it produces as well. In the first quarter, the company had planned for all of its Athlon chips to be built at its German fab.
However, when demand spiked as customers started buying Athlons rather than chips from Intel that cost 17 percent more, AMD added Athlon production in Austin to fill the orders.
The switch meant routing the silicon wafers on which chips are built to a different tool on the manufacturing floor, and downloading different software to print chip patterns and scrape away any extraneous material.
Engineers started to work with new software and procedures in 1995 and Ruiz has overseen the most of this work personally. Today, AMD can stop chip production at a few steps along the way and can switch to a new chip within two days.
That's a change from 1999, when it couldn't make enough of one chip to meet demand, let alone switch when one product proved more popular.
"We can't live in that kind of mode," says Behnke, the plant manager.
The first-quarter change to the Athlon in Austin helped AMD lift its sales CPUs for desktop PCs 17 percent over the December quarter. Intel, hurt by slack demand for PCs and servers, saw its PC processor sales fall 25 percent to US$5.13 billion during the same time period.
For Ruiz, the hardest part comes next: building future chips with SOI. He has, however, one factor strongly in his favor: About two-thirds of the 150 engineers on the project worked together on the Athlon, the CPU that's beaten Intel and AMD's first chip in years that didn't run afoul along the way.
SOI Server Chips
AMD plans to use SOI in server chips, as well as products for laptops and desktop PCs. Ruiz expects to extend the technique to top-notch systems now run with chips made by Sun, IBM Corp and Hewlett-Packard Co.
Intel derides the SOI effort, saying that the costs outweigh the performance gains. "We've got the fastest transistors without SOI," says Intel Senior Vice President Sunlin Chou.
Though his engineers say early work is promising, Ruiz admits that the finished product will be the real test. "You never quite know -- until you finally get silicon -- if you did what you thought you did," he says.
AMD workers agree that it's a hefty task to introduce chips in the second half of 2002 that use SOI, as Sanders has said they will. The effort is crucial, analysts say.
"Reputations are built very slowly and destroyed very quickly," says Brookwood of Insight 64. "If they were to fumble in some very visible way, people would say, `It's the old AMD again.' If Intel fumbles, no one says they'll never recover."
As high as the hurdles look, Ruiz says he'll win pieces of the laptop and server markets by reducing the amount of power that chips consume and by perfecting SOI. That will help him gain business-PC clients -- a bigger and more profitable market than consumer PCs.
If Ruiz fails to show consistent progress or if his bet on SOI fails, AMD will have lost time, money and perhaps its final shot at sustaining shareholder confidence.
RSS