IBM semiconductor alliance targets 32nm for metal gate fabrication

Tuesday 15th April 2008, 05:10:00 AM, written by Carl Bender

IBM announced Monday that the semiconductor research alliance led by the firm had successfully demonstrated significant process advancement through the incorporation of high-k/metal gate (HKMG) in test silicon at its East Fishkill fab. Anticipated originally for the 45nm node following a technology announcement in January of 2007, IBM and its partners will instead introduce HKMG fabrication at 32nm, with pilot production slated to begin in the third quarter of this year. Claimed performance gains from HKMG on 32nm - within the context of identical voltage to a 45nm control - include clockspeed increases of up to 35 percent and significant reductions in power.

In addition to IBM itself, 'Common Platform' partners Chartered Semiconductor and Samsung will offer OEM fabrication capacity featuring HKMG design, becoming the first OEMs to do so for the 32nm node. The process is further extensible to the 28nm half-node, and feasibility studies have concluded that HKMG fabrication should port successfully to the 22nm node in the future. The three are prepared to begin working with interested clients immediately to design chips around the process, and the partners are confident that the power/performance profile offered by HKMG will serve as a significant value-add differentiator to competing non-HKMG processes in the OEM market.

Other participating members of the research alliance named in the release include Infineon, STMicro, Freescale, and Toshiba. Toshiba is notable for being, along with Sony, a featured company in the original HKMG announcement. For its own part Sony has since scaled back its involvement in non-imaging related semiconductor production and left the research alliance, while maintaining its production interest in East Fishkill (Sony has invested hundreds of millions into the lines) and creating a new joint venture with Toshiba to manage semiconductor production for SCE. Sony also remains involved in the Cell research alliance; previous expectations were that the 45nm node might bring HKMG to Cell, revealed at ISSCC to remain SOI, and those predictions have now moved out to 32nm.

Another prominent company in the original announcement last year was AMD, absent as well from mention in the most recent release. Very much still a core member of the research alliance, AMD has claimed separately its preparedness to introduce HKMG at the 45nm node if need be. However, given the realities of its present fabrication capacity surplus and financial constraints, it is realistic to expect that AMD will wait until 32nm to introduce HKMG chips.

Discuss on the forums



Latest Thread Comments (2 total)
Posted by b3dusr on Tuesday, 12-Aug-08 08:46:36 UTC
32nm is good news, IBM and Chartered Semi have been working together on process technology since 2002, when they collaborated on the 90nm transition. Samsung came on board in 2004, and now the three companies form the core of an extended alliance that includes Infineon, Freescale, and AMD.

cpuFrequenz (

Posted by 3dilettante on Wednesday, 20-Aug-08 13:17:13 UTC
The significant (projected) improvements going from 45nm to 32nm HKMG might in part come from the transition from a 45nm that is "not too special" to a 32nm that "finally has what Intel had going on for years".

The upshot to this is that TSMC's 40nm process, in comparison to the supposed top dogs in the process race, might turn out to be "oddly competitive", given that outside of Intel the pace is slackening significantly.

As much as AMD likes to state the idea of having 45nm HKMG, I think we better see if they still exist when they hit 45nm non-HKMG, let alone 45nm HKMG or the distant 32nm horizon.

Then again, AMD might have to fall back on 45nm HKMG because it doesn't plan to transition to 32nm for quite some time after IBM and friends do. Once again, this is prefaced with the assumption that even AMD thinks it will be around (or will exist as chip designer that will even be creating high-performance CPUs) by that point.

Add your comment in the forums

Related News

Diving into Anti-Aliasing
RWT explores Haswell's eDRAM for graphics
ATI shoots a Bolt through its GPU compute stack
AMD releases CodeXL 1.0
Travelling in Style: Beyond3D's C++ AMP contest
Analysis of Ivy Bridge Graphics Architecture at RWT
RWT analyzes Kepler's architecture
Nvidia 680 GTX (Kepler) Released
Microsoft Releases C++ AMP Open Specification
Nvidia's 2x Guaranteed Program