By Bob Haavind, Editorial Director, Solid State Technology
A ConFab session on 450mm wafers provided a multifaceted look at a sharp split that has developed in the industry.
“We need a 30% cost/cm2 reduction every 10 years to follow Moore’s Law,” Peter Silverman, Intel Fellow and director of equipment technology strategy in Intel’s technology and manufacturing group, stated unequivocally. Mostly, he said, this has been achieved by wafer size increases along with the feature shrink, which provides the doubling of device density every 18 months to 2 years, the other part of Moore’s 1965 projection.
Scott Kramer, director of ISMI, showed projections that would fall short of the cost/transistor reduction goal even with a conversion to 450mm wafers in 2012, as the ITRS specifies. Kramer suggested even greater productivity enhancements will be needed to stay on Moore’s track.
But how will the industry pay for another costly wafer size transition? Mark Pinto, SVP and CTO of Applied Materials, showed that advanced technology R&D will cost toolmakers some $8 billion by 2010. Developing 450mm wafer-processing technology could double this, adding another $8 billion. Yet, since tools are becoming much more productive there would be fewer tools needed. This would not provide the payback essential to finance this major transition effort.
One alternative being explored, according to Kramer of ISMI, is called “300mm Prime,” an effort to increase the productivity of 300mm tools by 30%. This work would be compatible with a future 450mm transition whenever it comes, Kramer explained.
This concept needs to go much beyond just improving tools, which are already very productive but cost too much, suggested Arieh Greenberg, senior principal, productivity enhancement, Infineon’s Qimonda memory operation, from the audience. He believes entirely new factories must be designed with lower cost standard-platform 300mm tools, allowing faster setups and process transitions and operating much more efficiently. If the industry stays with 300mm, 30% improvements will not be enough — he believes 2x productivity improvement will be required.
Silverman of Intel pointed out that die cost has declined 3%-4%/year over the past 20 years, bringing similar reductions in average selling prices. This has driven the steady expansion of electronic markets. If the industry does not tackle the coming upward tilt of the cost reduction curve, the impact will start to be felt in the 2012 timeframe.
“You can do it with higher productivity 300mm equipment, or 450mm wafers, or with design innovation, but we *must* get more productivity improvements in 10 years,” he concluded. — B.H.