Intel smashes transistor limitations

Intel Corp. is trumpeting a technology breakthrough it says will lead to billion-transistor processors by 2007, resulting in computing power well beyond what's capable today.

The advance is a new kind of transistor, made under a new process and using as-yet-unspecified materials. It should help Intel solve current problems of power consumption and heat, as it creates smaller and faster transistors for future chips, says Ken David, director of components research for Intel's Technology and Manufacturing Group.

Still, an analyst notes that while Intel may have found one way around the problem, its competitors may have their own solutions by the time Intel implements the new technology, expected by 2007.

Dealing with leakage

The new technology should enable Intel to keep creating smaller, faster transistors for future chips, and keep pace with Moore's Law well into the next decade, David said in a conference call Tuesday.

Intel cofounder Gordon Moore posited that the number of transistors that can be placed on a sliver of silicon will double every two years. Moore said it in 1965, and Intel executives say it has remained true since Intel's 4004 processor shipped in 1971. That chip's transistor count: 2300. Current Pentium 4 processors contain about 55 million transistors.

A transistor is a simple switch that lets an electrical current flow from source to drain, David said. A gate electrode controls the flow of electricity through the gate dielectric. NMOS transistors are on when the gate is at high voltage; PMOS transistors are on when the gate is at low voltage.

Over time, transistors have become smaller, faster, and cheaper. As a result, however, the gate dielectric--made from a material called Silicon Dioxide--has become smaller and thinner, he said. The gate has become so thin (only a few molecules thick) that electrical current has begun to leak through it.

A good analogy is a tap that drips water when it is supposed to be turned off, David said.

The problem is, leaking transistors consume more power and generate more heat than they should. Unsolved, power and heat problems would eventually limit Intel's capability to create smaller and faster transistors, he said.

Industry and university labs have studied transistor leakage for almost 15 years, David said. An Intel team of researchers has worked on the issue for five years, and a result is this week's announcement.

Special K

To deal with leakage, Intel's team went searching for a replacement to silicon dioxide, David said. The substance needed to be thicker to reduce leakage, but needed to offer a high "K" value for good performance. (K is the dielectric constant of a material.)

Intel found its material, and the company won't reveal it. But transistors based on it work much faster and operate cooler, the researchers say. Specifically, they offer about 60 percent greater capacitance and a mere fraction of the current leakage.

However, problems are associated with the new material--namely, it doesn't work well with today's polycrystalline silicon gate electrodes. To solve the problem, Intel created a new type of gate electrode that uses a specific metal for NMOS transistors and a different one with PMOS transistors. Intel also declined to name its metal gate materials.

Intel isn't revealing the specifics of its new materials now because to do so would give away the company's competitive edge, David said. However, he noted the announcement goes beyond competition.

"This isn't about bragging about transistor speeds, but a way of showing that we've found a solution," he said.

The race goes on

It's important to note that while Intel is confident it's found the best solution to this problem, there may be other answers out there, notes Dean McCarron, principal analyst with Mercury Research.

"Lots of people are working on this because it's a well-known problem that must be solved," he says. "Intel believes they have found their solution, but it may not be the only one."

Still, Intel's announcement is important since a solution to the problem is a must if processors are going to continue to get faster and more powerful, McCarron says.

"I think there is a general presumption that things always get faster, cheaper and better," he says. "That has been true of processors, but its taking more research and development--and a lot more money--to build these new products," he says.

Join the newsletter!

Error: Please check your email address.

More about IntelMercury Research

Show Comments

Market Place