Wiring the world with Moore's Law

Intel on expanding Moore's Law Pat Gelsinger, CTO, Intel

Over the next 10 years, the Santa Clara, Calif.-based chipmaker will turn its research and manufacturing expertise toward driving down the cost and size of radios, optical networking equipment and other communications devices to the point where communication nodes will become omnipresent, Pat Gelsinger, Intel's CTO, said during a Thursday keynote speech at the company's developer conference here.

A year from now, for instance, laptops will contain chips that will let consumers more easily transfer from wired to wireless networks without a hiccup. Five years from now, radios will be so small they'll be tucked into a corner on a microchip, drastically reducing the cost and size of cell phones. Optical connections will be used inside computers to reduce power and heat.

"We could get it to the point where radios are built into every product we make, giving every device seamless, roaming connectivity," Gelsinger said. "You will see orders of magnitude of cost (decreases) through integration into silicon."

Integrating communications functions into microprocessors directly promotes Intel's overriding goal to become the world's primary supplier of chips to the Internet. Merging communications and computing chips means the company's products become more attractive to networking customers looking for networking chips with built-in intelligence and to computing manufacturers hoping to cut costs by eliminating the need to insert independent wireless cards.

Consumers will also increasingly wrap their lives around the Web if devices keep getting cheaper.

"We are seeking to define and extend the entire ecosystem," Gelsinger said in a separate interview. "That is the landscape we have launched ourselves into."

The trick for wiring the world boils down to using the basics of the computer chip business--Moore's Law and silicon manufacturing techniques--more pervasively and aggressively in the communications market, Gelsinger said.

Moore's Law postulates that engineers can double the number of transistors on a given chip every 18 to 24 months, primarily through shrinking the size of the transistors and other components. Along with increasing the power of chips, this evolutionary process drives down the price.

Although some obstacles exist, the rule will continue for years, Gelsinger said. "We stand here today more confident than ever in predicting another 10 years for Moore's Law."

The manufacturing techniques won't work exactly the same in all communications fields. Hypothetically, optical communications equipment, which transmits data via photons rather than electrons, may hit a finite shrinkage point because "a photon is only so small. You can't do half a photon," Gelsinger noted.

Cutting costs
The manufacturing techniques in the silicon world will nonetheless work to reduce costs. Currently, many optical components are customized and can be difficult to manufacture, as opposed to computer processors, which can be mass manufactured once the design is worked out. With integration, optical equipment that costs $10,000 now will sell for less than $1,000 in a few years, Gelsinger said.

Merging communications with computing will also help Intel from a business standpoint because it will essentially allow the company to use its existing resources to colonize new markets.

The company, in fact, has already begun to move high-end communications projects in-house. Currently, Taiwan's TSMC manufactures many Intel communications chips because these chips originally came from companies Intel acquired. However, Sean Maloney, general manager of Intel's Communications Group, said this week that many of the newer chips will be made on Intel's 130-nanometer manufacturing processes.

Research between the two fields overlaps, cutting costs again, Gelsinger said. Many of the compiler and tuning technologies used on Pentium processors, for instance, also work, with some adjustment, on the XScale line, the core chip architecture of the communications line.

Along with expanding the scope and direction of its research, Intel continues to expand its reach in the scientific world. Along with conducting experiments at its own research labs, the company performs research at Intel-funded "lablets" associated with the University of Washington; the University of California, Berkeley; and Carnegie Mellon University, and funds a number of other projects through grants.

Increasingly, an incremental amount of the company's research projects will migrate overseas. The company's research grants "are too heavily biased in the U.S. today," Gelsinger said during the interview. "The U.S. graduates about 50 percent of what U.S. industry needs."

Foreign engineers also need jobs. When the company opened offices in Nizhny-Novgorod, Russia, it received approximately 100 applications for every position. "We hired strongly qualified applicants with Ph.D.s for about one-fifth" the cost of their U.S. counterparts, Gelsinger said.