A good example would
be adjusting the etch conditions (maybe overetch) depending on the litho linewidth
measurements. This can be "dangerous".
One could "adjust" the etch process (perhaps % overetch) in order to compensate but by doing so one would acknowledge a problem and by "fixing" the problem a with subsequent processing step one might mask the problem rather figuring out the rootcause though fixing the problem on the fly by changoing percentage of overetch might work - but does it work when dealing with multiple manufacturing sites?
These kind of issues (at least to me) appear to become more complex and complicated when dealing with multiple manufacturing sites and considering we are dealing with several hundred (400?). I don't agree with following statemente because I believe AMD was never challenged to proliferate process to other AMD sites.
To me it's completely different story when dealing with one site compared to several manufacturing sites.
Critics say the Intel approach is overkill. At Advanced Micro Devices, Intel's biggest rival, engineers say the copy-exactly process stifles innovation. "Foolish consistency is the hobgoblin of little minds," says Thomas Sonderman, director of advanced process control for AMD. He describes his company's process as "copy intelligently." AMD, Sonderman says, prefers to innovate and introduce cutting-edge equipment right on the factory floor. "Fabs are slow-moving beasts. They are like aircraft carriers," he says. "We have changed the aircraft carrier into a destroyer. We can quickly change to what the market demands."
processing steps
Chip off the Old Block Intel ensures quality by using a slavishly identical process in all its plants.
(Business 2.0)
By Eric Pfeiffer
July 1, 2003
(Business 2.0) – If Intel's formula for success as a chipmaker were written down, it would fit nicely on an index card and possess an Einsteinian elegance: C+E+E=W+D, or "copy everything exactly equals world domination." ¶ The world's largest and most profitable seller of microprocessors, Intel has also become the most advanced manufacturing company in the business--thanks to what it calls its "copy-exactly" manufacturing process. Whether its chips are made in Arizona, New Mexico, Ireland, or any of its 10 other factories, they are made exactly the same way, with the same equipment, with workers performing the same tasks in the same order. What sets Intel apart from other manufacturers is its compulsive devotion to the process, even copying what may seem to be random parts of the process.
If a pipe that delivers chemicals to one of the chipmaking machines is 20 feet longer in one factory than in another, Intel will make it match (it will even match the number of bends in the pipe). If water quality is different among the factories, then "our purification system is meant to eliminate that difference," says Brian Harrison, Intel's vice president in charge of manufacturing. When one worker was found to be polishing the inside of an etching machine by wiping across the grain, he was asked to do it in the approved circular pattern instead. (It's less likely to drop specks of debris into the grooves.)
More from Business 2.0
How to tap your company's hidden network
Why it pays to invest in bosses who blame themselves
Bronfman's Warner Music remix
Fastest Growing Tech Companies
Current Issue
Subscribe to Business 2.0
So why does this matter outside the rarefied world of chipmaking? Any company could take a lesson from Intel on how to control millions of variables and potential problems. Figuring out the secrets of standardization is as key to Boeing as it is to Chrysler, Ford, and GM. Following the lead of Japanese automakers, GM has been so successful at standardization that it's pulling ahead of Ford and Chrysler to become the most efficient domestic automaker. Since 1997, GM has cut nearly nine labor hours from the building of a vehicle, adding an estimated $400 million to its bottom line last year alone. "We are real confident that [standardization] will take us to being the leader in quality in the entire automotive industry," says GM spokesman Brian Goebel.
Intel learned the copy-exactly process the hard way. At first the company's founders tried to convert the chip industry from a science to a down-and-dirty McScience--even privately calling the company McIntel, an homage to McDonald's. By the time the PC industry ramped up in the 1980s, Intel's chipmaking process had become so sloppy and inefficient that one of the company's billion-dollar plants didn't produce anything for a year, until its kinks were worked out. Craig Barrett was promoted to CEO and proceeded to tackle the problem.
It looked unsolvable at first. No other mass-produced item in the history of humankind is as complicated or as difficult to make as a computer chip. The average size of the latest transistors, for instance, is 90 nanometers, or one-thousandth the size of the period at the end of this sentence. Each chip has more than 100 million transistors, connected to each other by seven layers of stacked copper wiring.
Until Barrett stepped in, Intel's approach was end-loaded, which was typical of the industry. Quality control was achieved by trying to match output variables--making sure, say, that features on the transistor were the same thickness. Barrett's breakthrough was to focus instead on matching input variables, even on seemingly random things, such as the number of bends in the pipe. Hardly anything is left to choice except the color of paint--outside the building.
Matching input variables is a fundamental engineering principle. The Intel difference is isolating and controlling the phenomenal number of tiny inputs in the ever-shrinking chip. Everything from dust to water purity is scrupulously controlled.
The Barrett solution was to create one perfect plant and process--and then copy everything, even things that couldn't be precisely identified. It was the ideal solution for a digital world. "We aren't smart enough to know what matters and what doesn't, so we copy everything," Harrison says.
Critics say the Intel approach is overkill. At Advanced Micro Devices, Intel's biggest rival, engineers say the copy-exactly process stifles innovation. "Foolish consistency is the hobgoblin of little minds," says Thomas Sonderman, director of advanced process control for AMD. He describes his company's process as "copy intelligently." AMD, Sonderman says, prefers to innovate and introduce cutting-edge equipment right on the factory floor. "Fabs are slow-moving beasts. They are like aircraft carriers," he says. "We have changed the aircraft carrier into a destroyer. We can quickly change to what the market demands."
Still, many industry analysts believe that Intel has the best chip yield in the business. And it should, given how painstaking the process is. Take the story of its latest chip, code-named Prescott. Intel spent about two years in research and development in Oregon before transferring the process to its New Mexico plant. Factors that couldn't be changed, such as the difference between the barometric pressure in New Mexico and Oregon, were identified, and adjustments were made. During the two years, the entire process (taking more than 400 steps and employing 400 to 600 different machines) was gradually and precisely transcribed.
To alleviate human variables, Intel shipped 300 New Mexico workers to Oregon, where they spent a year working side by side with the R&D engineers, learning everything they needed to know. They picked up what Bruce Sohn, a manager at the New Mexico plant, calls "tribal knowledge"--information that Intel's most experienced employees know but may not have written down. "We want to copy everything--even the subtle things we may not even acknowledge that we do."
Finally, Intel "froze" the process, moving it to the new Albuquerque plant. Any future changes would need to go through rigorous peer review.
"It is a huge bureaucratic process," says Dan Hutcheson, CEO of VLSI Research, a market research firm specializing in the semiconductor industry. But it allows the world's largest chipmaker to mass-produce complex machines with a precision that even a decade ago would have seemed godlike. "It is," Hutcheson says, "one of the greatest achievements in the history of humankind." --ERIC PFEIFFER |
