Fast Forward: Another Leap For Intel

Posted 06/14/2011 at 11:40am | by Tom Halfhill

13

Comments

Print

Suddenly, like a plunging guillotine blade, Intel has severed any hope that competitors will match its chip-fabrication technology for years to come. Last month I observed that the rest of the industry was gaining a little ground on Intel by adopting high-k metal-gate (HKMG) transistors—only four years after Intel’s HKMG debut in 2007. But now comes Intel’s next big leap: tri-gate transistors.

Commonly known as FinFETs (finned field-effect transistors), these devices are a radical departure from the planar transistors used in integrated circuits for 50 years. Intel calls them the first true 3D transistors, but that description and the term “tri-gate” are potentially confusing. This isn’t 3D transistor stacking, which builds multiple layers of transistors on a chip. Nor does a tri-gate transistor have three electrical gates for controlling current flow.

Today’s 32nm transistor passes electricity underneath a gate in a flat plane. On the right is the new 3D transistor which increases surface area, lowers leakage, and increases density.

Instead, a FinFET is a three-dimensional structure that rises vertically above the chip’s flat silicon substrate. It looks like a tiny fin bisecting the gate structure. In contrast, planar transistors are flat devices etched into the substrate’s surface.

FinFETs can handle higher drive currents and switch between their on and off states at lower voltages. Chip designers can use those characteristics to reach higher clock speeds, or use less power, or achieve some combination of those advantages. FinFETs also leak much less current when the transistor is switched off.

Intel now has a fundamentally superior transistor, in addition to its 18- to 24-month lead in lithography. While Intel is moving into 22nm production this year, competitors are lagging a generation behind, just starting 32nm or 28nm production. Some companies don’t expect to have FinFETs until one process generation after Intel (14nm). Others have no FinFET roadmap at all. It adds up to a four-year lead for Intel—a huge obstacle for rival chipmakers to overcome.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.