The Finfet Play: Is This Crunch-Time?
The Intel Finfet play is like something from times past. Thirty or forty years ago chip companies would take out ads or hold press conferences saying stuff like: “Our ISO-CMOS junction-isolated 2.0 micron process delivers 30 per cent better performance than our best competitor while using half the power.”
This was how the industry differentiated its products. It was macho engineering stuff. Silicon transistors, Mesa transistors, the Planar process, MOS – each in their time revolutionised the chip industry and those companies which couldn’t make the transition were consigned to history.
And then CMOS came along, and everyone made the same CMOS structures, and the industry got boring from the manufacturing point of view.
But now Intel has re-opened the whole question of the importance of process. If its Finfet process can deliver in volume production what Intel says it can do – i.e. delivering up to 37% more performance without increased power or requiring half the power while delivering the same performance as planar transistors on the same process geometry – then every electronics OEM will have to think very carefully before using non-Intel chips.
Unless, of course, the industry can either match Intel’s prowess in manufacturing Finfets or come up with another transistor structure which delivers equivalent benefits.
If the rest of the industry can’t do that, and if Intel then moves aggressively into the foundry industry, then everyone competing in ICs on performance will have to use Intel for foundry.
The history of the industry suggests that none of this will happen. No one has maintained a fundamental, multi-year lead in process technology for a sustained period.
Militating against such a lead is the fact that the equipment industry makes its equipment for all and though Intel, TSMC, Toshiba and Samsung gets the stuff earlier than the others, the others do actually, eventually, get it.
And, at the same time, the great semiconductor research institutes like Imec will disseminate knowledge of how to make Finfets.
Indeed, last year, the President of Imec predicted 22nm as the process generation at which Finfets would first be used in anger. That is what Intel says it will do.
So, hopefully, others will master this new process technology and there will, once again, be competition on process.
If not, Intel will own the industry.