IBM and pals to start baking 28nm chips

Fast, efficient - just don't call it 'high-k/metal gate'

By Rik Myslewski in San Francisco • Get more from this author

Posted in Hardware, 16th April 2009 20:27 GMT

Free whitepaper – Reliability analysis of the APC Symmetra MW Power System

The march toward faster, more power-efficient microprocessors took another step Thursday when the IBM Technology Alliance announced the joint development of 28 nanometer process technology.

Alliance members include IBM, Singapore's Chartered, the recently-spun-off-from-AMD Globalfoundries, Germany's Infineon, Samsung, and Switzerland's STMicroelectronics. This global gaggle will work together to produce the process technology - already defined - that will be required to produce 28 nanometer high-k/metal gate silicon.

The Alliance, however, doesn't use the term "high-k/metal gate" to refer to that power-saving, leakage-reducing process technology - that's the terminology used by Intel, which first produced such chips in early 2007. The Alliance would prefer that you say "HKMG".

IBM and its partners were a bit late to the HKMG party, producing its first such silicon in April of 2008. While Intel started at 45 nanometer, however, IBM started at 32 nanometer. Now IBM and its partners - as well as Intel - have 28 nanometer in their sights.

Whatever you call it, HKMG - which replaces the polysilicon gates and silicon dioxide gate dielectrics in use since the 60s - is a major step forward in process technology.

As chips get smaller, their transistors are more prone to power leakage in the gate dielectric, leading to ineffiency and heat. An HKMG process can cut that leakage by five times or more.

Without HKMG technology, processes would be hard-pressed to keep up with Moore's Law, which states that chip complexities will double every two years. With HKMG, the Law remains in force - a least for a few more years.

According to the Alliance, early "risk production" of the 28 nanometer HKMG chips will begin in the second half of 2010. ®

Free whitepaper – Power and Cooling Capacity Management for Data Centers