Investing $7 billion dollars building fabrication plants.
Intel’s 45nm manufacturing process was introduced over a year ago now, realised in the Core 2 Quad QX9650 and was an impressive feat of engineering – getting transistors so small is a challenging task! This year will see a similarly notable process change, the shrink to 32nm which Intel has now started demonstrating with working CPUs.
The process is similar to that used to create Intel’s 45nm processors, though Intel stresses that this second generation is much refined from its predecessor. Fundamentally, though, the shrink in transistor size is facilitated by using the same combination of a high-k dielectric material and a metal gate. The refinements Intel has made include a reduction in the dielectric oxide thickness from 1nm to 0.9nm, while gate length decreases from 35nm to 30nm.
These numbers may seem small, but at this end of the scale they make a notable difference; Intel reckons its 32nm transistors gain a 22 per cent performance improvement over their 45nm brethren. Source-to-drain leakage has also been reduced with this generation of manufacturing process. In real terms, transistors require less power to turn on and off, which should lead to more power efficient CPUs.
Intel will be investing some $7 billion into its 32nm manufacturing technology over the course of this year. The money will find its way to 32nm production facilities in New Mexico, Arizona and Oregon ensuring that Intel is ready for mass production of 32nm chips come Q4.