RealTime IT News

Intel Co-Founder Wants to Delay 'Forever'

SAN FRANCISCO -- When Intel co-founder Gordon Moore postulated the size and speed of future chips, the thinking was that things would eventually slow down because of the limits of silicon.

Now the 74-year-old Chairman Emeritus isn't so sure.

"No exponential is forever, but we can certainly delay 'forever,'" Moore said appearing before the IEEE International Solid-State Circuits Conference here. "Exponentials tend to distort what we are capable of. This is the largest value-add industry that I know of because we start with sand and produce computer chips. There is still a lot to do and there are still some clever ways of achieving our goals. There was a time when you couldn't get a gigabyte on the highest IBM machine."

Overall, the chip industry has maintained an average growth of 80 fold every year and Moore sees no sign of slowing down in either the amount of transistors being shipped or the incredible shrinking size of the chips. For example, current conventional transistors are produced at the 130-nanometer (nm) and 90nm process level with 60nm due out in 2005. Given Moore's estimates of 2 or 3 years between generations, 45nm and 30nm processors should be readily available by 2010.

"Below 30nm it's not clear which direction we will go because there is always some type of catastrophe in the future... always some type of challenge," Moore said.

Part of the solution is a necessary step in lithography to print and etch lines on semiconductor wafers closer together.

In that stead, Moore said Santa Clara, Calif.-based Intel is invested heavily in Extreme Ultraviolet (EUV) lithography, or EVU processes. The technology was spawned from Regan-era Star Wars program in a partnership between Sandia National Laboratories and Lawrence Livermore National Laboratories in California. The technology has had a profound number of transistors on a die, rising from 10 in 1960 to an estimated 1 billion sometime in the next two years.

"I thought that 1 micron would be as best as we could go, then we moved Ultra Violet light and now we're sub 10 micron and looking at printing lines at even smaller sizes," Moore said.

The other motivating factor in reducing size and power output is price. Year after year the price of semiconductors has dropped from about $1 in 1968 to 2 tenths of a dollar today. DRAM chips are even lower with 250 transistors going for a dollar.

Moore admitted that early on, Intel's strategy was dropping the price to spur usage of its processors.

"You might think integrated circuits were easily accepted, but it was a hard sell," confessed Moore. "The circuit design people and the reliability people balked at how much it cost to build their own chips. We came in with a lower price and told them we would make it cheaper."

More pressing however is the industry's concerns over processor power in terms of active production and leakage.

"Currently the amount of power being put out by semiconductors is about at hot as a light bulb," said Moore. "I don't want a kilowatt in my lap. That would be uncomfortable. If power is going to have to come down we're are going to have to address this."

With the problem of leakage, Moore says the key is processor supply voltage. In the beginning, the standard was 12 volts with the most hovering near 2 volts.

"Again this can't go on forever just to overcome some of the noise problems. I suspect 1 volt would be the limit, but I have been wrong before," said Moore.

Intel is scheduled to present some 12 separate papers to the IEEE covering ways to decrease the amount of power needed to run computers while minimizing leakage.

The various technologies include a 5GHz Floating Point Multiply-Accumulator (MAC), new "Sleep" Transistors, an improved Multiphase Clock Generator and a 1.5GHz third-generation Itanium Processor.