Inside the IBM/Chartered/Samsung chipmaking partnership

by Ed Korczynski, Senior Editor

After leading much of IBM’s technology development over a multidecade career, Walter Lange recently joined Common Platform Alliance partner Chartered Semiconductor Manufacturing as EVP of technical marketing. In an exclusive interview with WaferNEWS during the recent Chartered Technology Forum (Sept. 28), Lange offered frank and insightful comments on Chartered’s R&D strategy and its manufacturing goals, the business reality of competing with partners, and some amusing stories of what it takes to prove a new technology really works.

Lange is now strongly focused on foundry business strategy given his new role at Chartered. While acknowledging the current reality of TSMC commanding >50% of global foundry business, he envisions a near-term future where designs didn’t have to be “T-like.” “Foundry isn’t foundry anymore,” Lange said; “solutions are what you need.”

Chartered believes its constellation of labs, fabs, design tools, and system integration expertise will provide greater value than what rival TSMC can gain from the Crolles2Alliance with partners Philips (now NXP), Freescale, and STMicroelectronics. The perspectives of IBM and Infineon provide systems-level perspectives on both design and manufacturing which is essential for cost-effective nanometer-era IC production. For the right price, the IBM/Chartered/Samsung chipmaking alliance offers everything needed to bring an IC design in an advanced 65nm node IC to high-volume packaged chips.

“Because of the way the partnership is built, there is a lot of flexibility,” said Lange, adding that the common platform allows customers “to work with their strengths.” A customer with internal design capability, for example, might do its own design-rule-checking and lithography-rule-checking to create a final GDSII file, and use the platform strictly for manufacturing. If, however, a company has deep pockets and tight market windows, then the platform can do both custom design work and chip-package integration to minimize the risk of getting first-time-right silicon. “How do you minimize risk?” mused Lange, “There’s no one correct answer, since every company has its own strategy and approach.”

The Microsoft Xbox chip is an example of the alliance’s ability to attract big customers. Leveraging its systems architecture and design prowess — “there’s only one company in the world that can they do a custom microprocessor,” Lange commented — IBM did the basic design as well as DFM design services to optimize the GDSII file for making the masks, which can be run on any one of the alliance partners’ fabs. Chartered assured Microsoft that volume production would be available regardless of any possible regional issues (e.g. natural disasters), and at market prices in high volumes. “They would not have been able to make their December 2006 shipment schedules if there had not been multiple fabs,” informed Lange.

The highest-level business strategy for the alliance is also clear — the partners all collaborate, but also compete. They serve a huge business with a big pie, so there’re a lot of good size slices to go around. By providing GDSII-compatibility across multiple fabs, customers can move manufacturing from one company to another with the expectation that yield should be at least comparable. “If you want to manage your own destiny, you want to manage multiple sources,” said Lange. “That ensures that you get market-based pricing, instead of being held by an uncomfortable part of your anatomy.”

Technology development for the alliance continues to occur at IBM’s East Fishkill campus. Lange noted that the engineers are New York residents, treated as one team by the management structure — “a manager of a dry etch module might have ten people total from the different partners,” he explained.

The common platform alliance claims unexpected benefits from having separate teams working on ramping three separate “copy-smart” fabs around the world. “Copy-smart” means that upgraded versions of manufacturing equipment can be used, as long as the essential process on the wafer is not changed. Starting three lines in parallel allowed manufacturing learnings to be shared, such as defect-signatures and root-causes-analyses, and yields ramped much faster than anticipated.

Lange noted that most of what the industry now categorizes as “DFM” has been done quietly internally by lBM for generations, and that much of it is obvious work that must occur to manage design complexity. Mused Lange: “How stupid do you have to be to not do DFM?”

Looking back at IBM’s history of technology innovation, Lange shared an anecdote about doing whatever it takes to prove a concept in R&D. The earliest chemical-mechanical planarization work at IBM, where it was first secretly conceived as a crucial new manufacturing concept, used a crude mechanical spinning plate without a sophisticated pad or slurry dispense mechanism. Since centrifugal force drove off the slurry, two men in raincoats and face-shields sprayed fresh slurry onto the spinning plate using automobile pressure-hoses. That spirit of innovative continues to create innovative technologies — and now there are partners to help hold the hose. — E.K.


Easily post a comment below using your Linkedin, Twitter, Google or Facebook account. Comments won't automatically be posted to your social media accounts unless you select to share.