from WaferNews, V12n38, October 10, 2005
by Dr. Paula Doe, Contributing Editor
by James Montgomery, News Editor
While the common 90nm-65nm platform announced by IBM, Chartered, and Samsung on Sept. 22 aims to offer users the clear benefit of multiple manufacturing sources for a single design, and help fabless design houses better match the IDMs in designing for manufacturability, it’s less clear how much of an impact such cooperative partnerships of big chip manufacturers will actually have on suppliers. In fact, after three years, the partnership’s most impressive accomplishment actually may be that the participants are still happy with each other.
“It’s really interesting that they have been able to pull it off,” echoed Risto Puhakka, president of VLSI Research. “One of the big hurdles in a two- or three-way partnership is that the parties need to feel that they are getting more out of being in a partnership than being alone.”
One very big endorsement to the alliance, Puhakka noted, is that Samsung has discontinued its own independent process development efforts in logic, including design, to align them with the consortium. Gartner Dataquest analyst Dean Freeman added that Sony and Toshiba, working with IBM on Cell technology, could potentially apply the common platform for different applications, e.g., set-top boxes and gaming systems.
Chartered officials showed a slide of what the alliance hopes will be another benefit for participants: combined capital spending budgets far exceeding single chipmakers such as Intel and TI, or other alliances such as Crolles. “The opportunity for suppliers is to communicate with the members as a group,” said John Martin, Chartered’s VP of strategic alliances. “The potential is that the common platform and the common development partners are involved in tool selection for development, so the potential for identical or very similar implementation is reasonably high. [Working with the alliance] provides attractive capacity for development and the possibility deployment by multiple users.” He noted, however, that the “copy smart” strategy allows some room for adjustment, for example when there’s a more recent generation of the tool, or when a particular supplier has much better support in one region of the world.
The IBM/Chartered “copy smart” philosophy means that the bulk of the equipment set is probably identical, although there may be small differences in noncritical equipment, said Freeman. “For equipment suppliers, that’s a bad thing: fewer places to show your wares, and fewer people to buy them,” he said. There will be more consortia-style projects due to chipmakers’ desire to balance R&D dollars and minimize risk, he predicts, creating further division among semiconductor equipment manufacturers who will need to get their equipment ready for R&D even sooner. “If you’re serious about being in the 45nm race, you’d better have a piece of equipment that is working on 45nm right now” either in place at an R&D fab or running internal demos, Freeman said. Suppliers that can’t or won’t work with large consortia “will find business with niche players or small foundries,” but that’s a limiting proposition since the top 10% of chip manufacturers account for 80% of industry capex.
Puhakka noted that partnerships really getting leverage on their capex will require a much higher level of collaboration. “That means going to ASML and saying they need four immersion steppers, shipped on specific dates and at specific prices,” said Puhakka. “For now they may need four, but not at the same time. That kind of breaks leverage out of the whole deal.”
The partners apparently have managed to work together to ramp up 90nm production with impressive efficiency. Chartered’s Martin noted that while Chartered shipped its first 90nm revenue product in the June quarter, it projected 22% of its total revenues would be from 90nm in the September quarter. By contrast, the company’s initial revenues from ramping 130nm production increased at only about 3% per quarter, from 2% of revenues in 1Q03, to 10% in 4Q03.
Besides involving common development team members at both IBM’s Fab 323 and Chartered’s Fab 7, using common SPICE models, and focusing on precise matching of physical and electrical characteristics, the partners credit PDF Solutions’ tools used to identify process defect issues. “We identified and compared the best and worst areas in both fabs,” says Martin, “which lead to very fast improvement. As far as we know we’re the only one who has done this — or at least the only public case.”
Chartered has separately licensed significant portions of advanced process control technology from AMD, whose joint development agreement with IBM focuses on SOI, not this bulk CMOS technology. But these processes are still being implemented, so they haven’t yet impacted the ramp.
Overall, the partnership has been particularly good for Chartered, which “jumped from 130nm now down to 65nm very quickly as a result of this relationship,” Freeman said. The ability to manufacture leading-edge devices for IBM, as well as to have process capability to do that as a foundry for other fabless companies, “puts them on par with TSMC,” he said, “and maybe gives them a leg up because they have an active partner for 90nm-65nm technology, where TSMC has to develop the technology themselves or get it through Crolles.”
Meanwhile, for IBM, the partnership has created more capacity and flexibility for its foundry business and a second source for manufacturing ASICs, as well as any other internal IBM devices, the analysts said. The partnership also helps them pay for R&D. — P.D. and J.M.