The world's leading source of technology news and analysis
Search Spectrum IEEEXplore Digital Library Submit
Font Size: A A A
IEEE
Home [Alt + 1] Magazine [Alt + 2] Bioengineering [Alt + 3] Computing [Alt + 4] Consumer [Alt + 5] Power/Energy [Alt + 6] Semiconductors [Alt + 7] Communications [Alt + 8] Transportation [Alt + 9]

The Amazing Vanishing Transistor Act Continued By Linda Geppert

emailEmail PrintPrint CommentsComments ()  ReprintsReprints NewslettersNewsletters

Racing to the limits

Although some pundits have predicted that the evolution of semiconductor technology to smaller dimensions will slow down as dimensions shrink, things are in fact speeding up. The International Technology Roadmap for Semiconductors, published periodically by the Semiconductor Industry Association (SIA, San Jose, Calif.), recently revised its projection for the 2003 technology node from 100 nm to 90 nm. "Technology node" refers to the set of processes needed to print the smallest feature, which would be approximately 90 nm. In high-end processes, gate length may be selectively etched down to about half this minimum feature size.

True to the 2001 Roadmap projections, many foundries, including Intel, TSMC, Philips, IBM, STMicro, Motorola, and LSI Logic, are gearing up to start volume production of 90-nm processes in 2003. Intel Corp.'s prototype 90-nm process, being brought on-line at its Beaverton, Ore., facility, has already produced a fully functional 52Mb SRAM with transistor gate lengths of 50 nm and SRAM cell sizes of just 1 um2, or roughly half the cell size of today's most advanced SRAMs [see figure, "The Amazing Vanishing Transistor Act"].

The downscaling will continue. According to the SIA's roadmap, high-performance ICs will contain by 2016 more than 8.8

billion transistors in an area 280 mm2—more than 25 times as many as on today's chips built with 130-nm feature sizes. Typical feature sizes, which are also referred to as linewidths, will shrink to 22 nm, less than one-fifth the width found in the current generation of ICs, and power dissipation on high-performance microprocessors will double its present value, requiring more elaborate heat sinks. Some of the extra power consumption will come from gate-to-substrate and source-to-drain current leakage that will grow larger as channel lengths scale down to a few tens of nanometers.

Almost nine billion transistors on a chip may sound like a pipe dream, but scientists are already devising ways to make it a reality. Advances must be made on all fronts, including the chip manufacturing process, circuit architecture, and design methods. But no area is more essential to the future of semiconductor technology than the transistors used to build CMOS circuits [see sidebar, "The Gate Rules" for some basics of CMOS transistors].


« Previous Page 2 of 7 Next »
emailEmail PrintPrint CommentsComments ()  ReprintsReprints NewslettersNewsletters