Milton Friedman, the University of Chicago economist, certainly would have been happy to hear about the high-k race between Intel and AMD/IBM.
Friedman, who died last year, was a brilliant proponent of competition, personal free choice, and less government interference in business. He also hated monopolies, so I suspect in his later years he cheered for AMD’s design and factory control innovations while admiring Intel’s process technology breakthroughs and manufacturing efficiencies.
There is little doubt the AMD/IBM team has its work cut out for it, following the Jan. 26 announcements from both the Intel and AMD/IBM camps that they will use high-k and metal gates at the 45-nm node.
Consider the 2006 IEDM paper from the IBM Systems and Technology Group and AMD, describing their 45-nm SOI technology. The paper doesn’t mention the gate oxide at all, emphasizing advancements in immersion lithography and interconnects. One must assume that high-k and metal gates were not used in these transistors.
That no-high-k assumption is based on the numbers.
The IBM/AMD paper last December reported that their 45-nm transistor had an AC drive current -- at 1 V operation and an Ioff of 200 nA/micron -- of 840 micoAmperes/micron for the PFET and 1240 mA/micron for the NFET.
At the 2005 IEDM, Intel reported on their 65-nm process. At 100 nanoAmperes/micron of off-current at 1V operation, the PMOS transistor had a drive current of 710 microAmperes/micron and the NMOS had 1,210 microAmperes of drive current. (I don’t know what the numbers would be for Intel if it used the 200-nanoAmperes of off-current that IBM and AMD used in their 45-nm paper.)
If you take Intel senior fellow Mark Bohr at his word that Intel will achieve a 20 performance gain in its 45-nm transistors compared with its 65-nm transistors, Intel should be in the ballpark of 852 microAmperes/micron for the PFET and 1,452 microAmperes/micron for the NFET. (My guess is that less than half of that 20 percent gain comes from bringing in high-k and metal gates, and the rest from enhanced strain and other techniques. Also, I realize that these are self-reported numbers that can be taken with a grain of salt for all concerned.)
Surely, if IBM and AMD bring in high-k and metal gates as they say they will, they can match Intel’s transistors by mid-2008.
The real question is timing. At a demonstration for the press at Intel’s headquarters last week, Intel demonstrated five working microprocessors, running in server, desktop, and mobile systems, using operating systems from Sun, Apple, Microsoft, and Linux. Those working systems demonstrate that Intel is well-prepared to begin manufacturing 45-nm MPUs in the second half of this year, and to have products in the hands of end users by the end of this year.
For AMD and IBM to hit their mid-2008 target for their 45-nm products, they will need to burn the midnight oil to bring high-k into working processors. They are probably used to that, given the success of AMD’s Opteron products the last few years, and the increasing support for the Fishkill alliance.
Milton Friedman was right. Competition benefits all of us, and there is no more intense competition than the 45-nm race between Intel and IBM-AMD.
About weQuest: weQuest's are written by G Dan Hutcheson, his career spans more than thirty years, in which he became a well-known as a visionary for helping companies make businesses out of technology. This includes hundreds of successful programs involving product development, positioning, and launch in Semiconductor, Technology, Medicine, Energy, Business, High Tech, Enviorntment, Electronics, healthcare and Business devisions.