Barrett: No end in sight for Moore's Law

The momentum will be kept up first through conventional manufacturing processes, then for many years after that by other technology, he said in a keynote speech at the Intel Developer Forum here.

Barrett predicted that traditional chipmaking technology will permit features as small as 5 nanometers--about the width of 50 hydrogen atoms--to be used on processors.

"We can see how to do this down into the 5-nanometer range," Barrett said. "Beyond that, lots of leakage currents and things like that get in the way. But every time we seem to get into a roadblock, the bright engineers...seem to circumvent that problem."

The future of Intel and the computing industry in general depends in large measure on the ability to pack more circuitry components, called transistors, ever more tightly onto a slice of silicon. To do that, the size of chip features must be shrunk.

Intel today is preparing to introduce processors with features measuring 65 nanometers, or billionths of a meter. Company engineers have forecast the feasibility of 5-nanometer manufacturing processes before. But the public declaration of the chipmaker's top executive carries more weight.

"He was willing to extend the planning horizon to 5 nanometers," Peter Glaskowsky, an analyst at Envisioneering Association, said. The rest of the computing industry will be able to factor that into their plans, he added. "That means everybody gets to stay on the treadmill."

Barrett showed photographs of transistor prototypes built with features measuring 65, 45, 32 and 22 nanometers. Glaskowsky said current expectations are for new processes with features of 15, 10, 7 and 5 nanometers. However, it is expected that the two-year pace of each new manufacturing process will slow, he said.

Conventional chip manufacturing processes use a technology called complimentary metal oxide semiconductors, or CMOS. It's not clear what technology will replace that to create even tinier transistors, but Barrett mentioned three options: quantum dot, polymer layer and nanotube technology.

"I don't think the world has decided what is the replacement yet," Barrett said. "Meanwhile, it's full blast ahead with the standard Moore's Law."

Moore's Law is the 1965 projection by Intel co-founder Gordon Moore that the number of transistors on a chip will double about once every two years. "I've told Gordon I plan to help him celebrate the 50th anniversary of Moore's Law in 2015, and hopefully the 60th anniversary in 2025," Barrett said.

Barrett plans to leave Intel's chief executive post in May, and will be replaced by President Paul Otellini. The company is in the midst of a reorganization and an effort to prevent a repeat of 2004, which was plagued by troubles.

Intel chips are widely used in personal computers, but the company has had a hard time expanding into the cell phone market, Barrett acknowledged. He said progress will begin this year.

"We haven't given up, and we are kind of doubling down in that space," he said. "We have done well in the PDA (personal digital assistant) space, but we have done less well in the cellular space, and we expect to see a number of design wins later this year."

Another sore point for Intel has been the Itanium processor family, once aimed at the entire market for powerful networked computers called servers, but now designed just for high-end systems. The adoption of Itanium has been slower than Intel had hoped, but the company still has long-term plans for the chip.

Intel will keep Itanium separate from the more widely used x86 chips such as Pentium and Xeon, Barrett said. "I don't see a great need to merge the two architectures," Barrett said.

Itanium will compete increasingly with Power chips from IBM and Sparc chips from Sun Microsystems and Fujitsu, Barrett added. Hewlett-Packard, which codeveloped Itanium, is the chip's biggest supporter. Intel also is pushing into wireless communication technology. Barrett reiterated his company's support for WiMax, which can blanket a region as large as San Francisco with high-speed wireless networking abilities.

"I think WiMax is going to be a disruptive technology that's going to change the way we think of mobile connectivity," including mobile phones, Barrett said. "Hopefully, it will get us out of the half-assed broadband capability we have today in the U.S."

WiMax technology currently awaits adoption by network service providers. "Hopefully, toward the end of 2005 or 2006 you're going to see massive commercial rollouts of this capability," Barrett said.