The 2011 International Electron Device Meeting is held annually to describe the state of the art in transistors and switching circuits. This year’s Washington meeting is exploring a variety of options for continuing Moore’s law for as long as is feasible. The keynote speaker, Intel Senior Fellow Mark Bohr, noted that traditional scaling ran out of steam in the early 2000s, and that the semiconductor has had to repeatedly come up with innovations in order to keep semiconductor scaling viable. At the 90nm node, Intel introduced Strained silicon, and 45 nm microchips featured high-k metal gates. The 22 nm generation, which is now in volume production at Intel, is made using tri-gate transistors. Intel is particularly proud of the tri-gate advance because tri-gate transistors require less doping, leading to better performance, reduced leakage, and less transistor variation. Intel is confident that these tri-gate transistors, which can be used in everything from mobile devices to supercomputers, will give it a competitive advantage over competitors using conventional planar transistor technology,.
These innovations have served to substantially reduce leakage, and have allowed Intel to envision producing .7 volt microchips. For the 14 nm generation and beyond, Intel is examining a number of approaches, including III-V compounds such as gallium arsenide and indium phosphide. These compounds have significantly greater mobility than silicon, but could increase manufacturing costs. Other options that Intel is exploring include “wraparound” nanowire devices, tunneling FETs, 3-d chip stacking, and graphene FETs. Graphene transistors have extremely high switching speeds, but have no bandgap. Unless ways are found to give graphene a bandgap, graphene will not be suitable for digital devices. Whatever options are introduced, Bohr made it clear that Intel is now more focused on switching efficiency and power per watt than raw switching speed. In 1996, Intel CEO Andy Grove made a prediction that in 2011 microprocessors would operate at 10 GHz. Given the industry preoccupation with leakage/power issues, such a microprocessor may never see the light of day.