Pages: p. 82
A string of record-breaking results turned in by leading semiconductor companies recently have failed to dispel long-term fears of an industry downturn. The sector's benchmark, Philadelphia Semiconductor Index, fell from its year high of 560 to 416 on 30 July, and the mood of uncertainty was showing no sign of lifting.
China's largest chipmaker, Semiconductor Manufacturing International, followed the trend set by other semiconductor companies in reporting record second-quarter sales and profits.
Sales jumped almost 200 percent in the same period last year to (US) $221 million, while profits surged to a record $34 million. Earlier, industry giants Intel, Texas Instruments, Taiwan's TSMC, and Hong Kong's Solomon Systech all reported sales and profits that either matched or exceeded analysts' predictions. None of these had been enough to reassure nervous markets.
The malaise affecting the industry was first heralded by a Merrill Lynch report on 12 July in which the investment bank shifted its position from "overweight" to "underweight" and warned that earnings might have peaked for the sector and the recovery would end earlier than analysts forecast. A 15 percent rise in inventory at market leader Intel gave rise to Merrill Lynch's gloomy assessment. It promptly slashed its 2005 growth forecast for the industry from 16 to 6 percent.
In the latest misstep for the world's largest chip maker, Intel said a faster version of its flagship Pentium 4 microprocessor will not be available by the end of the year as previously promised. The company said that the 4-GHz chip will not ship until the first quarter of 2005. Currently, the fastest model runs at 3.6 GHz. It's been a rough year for Intel products, even though profits nearly doubled in the second quarter and it was on track to set record sales in the third quarter. Its current Pentium 4, unveiled in February, has been criticized for not raising the performance bar over previous processors.
Challenging laser fuse techniques, IBM announced a chip morphing technology based on electromigration, which it says can allow a new class of semiconductor products that monitor and adjust their functions without human intervention.
Called eFUSE, the technology is part of a built-in self-repair system that constantly monitors a chip's functionality. eFUSE works by combining software algorithms and microscopic electrical fuses, as opposed to laser fuses, to produce chips that can regulate and adapt their own actions in response to changing conditions and system demands, according to IBM.
If an imperfection is detected, eFUSE initiates corrective actions by tripping inexpensive, simple electrical fuses that are designed into the chip at no additional cost, IBM explained. The activated fuses help the chip control individual circuit speed to manage power consumption and repair unexpected flaws.
If startup Cambrios is right, semiconductors and other computer parts in the future won't be built. They'll be bred.
The Palo Alto, Calif.-based company is using methods that will allow researchers to build semiconductors or other components by combining inorganic substances such as cadmium sulfide with a vast library of genetically engineered organisms. Cambrios is formerly known as Semzyme.
In the vast majority of situations, combining a metal with a living virus or bacteria won't result in a breakthrough, but occasionally the chemical interaction between the metal and a protein from the organism produces elegant—and potentially commercially attractive—films or crystals, said Cambrios chief executive officer Mike Knapp. A seashell, after all, is chalk that has reacted with specialized proteins.
"This is the way evolution works. You try lots of stuff and see what works," he said. "Proteins can manipulate things. We wouldn't survive as humans if our proteins didn't manipulate things atom by atom."
Researchers have long discussed techniques for adopting processes found in nature, but practical advances in nanotechnology appear to be inching closer to reality. United Kingdom–based Nanomagnetics is examining ways to make tiny, uniform memory cells out of proteins. Japan's Matsushita is conducting similar research. Other companies are looking at ways to harness photosynthesis for energy production.
A research team from Michigan-based universities says it has succeeded in integrating the last two components needed to create a one-chip wireless transceiver.
"Our research group picked up the challenge to integrate the last two off-chip components onto a wireless transceiver," said Michael Flynn, head of the wireless-interface group at the Wireless Integrated Microsystems Engineering Research Center (WIMS ERC) at the University of Michigan (Ann Arbor).
"Thanks to Kamal Sarabandi, we have demonstrated a Zigbee (2.4-GHz) wireless link using our 1-centimeter-square slot antenna and thanks to probably the world's foremost expert on RF MEMS (microelectromechanical systems for radio frequencies), Clark Nguyen, we have also developed a wineglass-like resonator to replace the off-chip quartz crystal.
"Now all the wireless components can be on one chip—enabling everything from hearing aid-sized cell phones to smart dust," said Flynn.