Intel trying to cool concerns over power consumption in its ever-shrinking processors by changing the way it makes the transistors inside them.

The company Monday outlined a long-term product roadmap that will see Terahertz transistors hit the market by the middle of this decade. The Terahertz products will feature an ultra-thin layer of silicon on top of an embedded layer of insulation to switch them on and off faster while reducing leakage. Right now if the transistor is not doing anything, it still leaks current to the substrate, which means that things like laptops or low-power devices are drawing power, even when they’re in a low-power state.

Intel will also replace silicon deoxide on wafers with a high-k gate dielectric the company says will allow a factor of 10,000 reduction in current leakage across the gate dialectic. The transistors will include thicker source and drain regions.

Doug Cooper, country manager for Intel of Canada, said the moves come as industry experts worry about high operating voltages and current leakage as more transistors are packed into chips to keep up with Moore’s law.

“Even as we made transistors smaller — 10, 20, 50 nanometres — the amount of power consumed just through these leakage currents was just going to make it very difficult to take a very high-performance product and put it in a portable device,” he said.

Processor generations typically last two to three years, and since this process is coming around 2005, it should carry Intel through the decade, Cooper said. “We’re looking at transistors that are operating now at this point in the range of a trillion to 2.5 trillion cycles per second, which is 10 to 20 GHz, processor-wise.”

Though there are different needs for notebooks, desktops and servers, Cooper said he expects to see the Terahertz transistors introduced across the entire line. “Mobile’s an obvious place where you want to reduce power consumption, but even in desktops, when we get to a 10 GHz processor, we’d like to maintain a flat power curve,” he said. “So that we’re not looking at any kind of exotic cooling (system).”

Dean McCarron, principal analyst with Mercury Research in Scottsdale, Ariz, said there are still other issues, raw clock rate for example, which aren’t going away anytime soon.

“There’s a fundamental physical limitation on how fast an electron can travel,” he said. “That will set a cap on the broad clock rate on a part. The solution to that is doing deeper and deeper pipelines where you’re doing small amounts of work in each incremental step, and therefore you’re dealing with a much smaller number of transistors in a given stage.”

Intel has long worked with major universities to examine ways to build future processors, including the possibility of molecular-based transistors using nanotechnology. Cooper said these efforts are closer to fruition than people think.

“If you look at the geometries we’re talking about, the gate oxides on these are as thin as three atoms,” he said. “We’re laying down a layer of atoms at a time, which really is nanotechnology.”

Whether or not the Terahertz design solves the power problem, McCarron said the processor industry is already working faster than the industry demands.

“The limitations are a moving target,” he said. “If you draw the line and say where we should be five years out and try to solve that problem with today’s technology, it can’t be done. It’s been that way for the last 20 years.”

Share on LinkedIn Share with Google+