Shrinking architectures are always news, and the industry has lately been seeing many 2xnm and 300mm advances, adoptions and initiations of mainstreaming at these next generation levels (see for example this recent discussion on 22nm NAND flash at 300mm wafer sizes from MarketWatch).
The news this week from Intel regarding their tested and demo'd 22nm transistor, means that the much awaited, much discussed 3D structures are now (finally) coming to the more mainstream production level. Intel's new, 22nm Tri-Gate transistor will be used in their high-volume Ivy Bridge Core processors beginning the end of this year (see this review from EETimes Europe).
The 3D transistor breakthrough is not an insignificant one and is at the heart of the importance of this announcement; the 22nm architecture is not the surprising part, as we've already seen a number of manufacturers break into this level with production roll-outs expected during 2H11 (see this discussion of fabs and shrinking nanometers from MarketWatch).
CMOS has been rooted to the planar (aka 2D) level without mainstream movement into 3D, despite serious R&D, discussions and designs having been brought forth over the past couple of years. Intel has pushed up the gate, literally, with this new structure, and therewith others must follow or fall behind. No longer will there be the luxury of comparing and perfecting 3D designs only in the R&D arena.
The specifications of Intel's new, 22nm 3D Tri-Gate transistor include (see this discussion from ElectroIQ):
- 364Mbit array size
- >2.9B transistors
- 3rd-generation HKMG
- The same transistory/interconnect features as on 22nm CPUs
- Only 2%-3% add to process costs (vs. 10% for FDSOI)
Intel, however, has 'only' demo'd this exciting transistor breakthrough to the laptop, desktop and server series at this time (see this review of what some analysts are saying by ElectroIQ). The big question reverberating the semiconductor industry news since the announcement, is, 'what about mobile?!' (see this FT article and this one from ComputerWorld, for example) That's a critical question. It seems from Intel's site that a mobile play is not omitted from their market goals for this transistor, but specific information is not provided. With the 50+% power reduction at constant performance, or 37% performance increases at low voltage, the application target for mobile is a no-brainer and cannot be far off.
On a broader market strategy play, to maintain their dominant role while also making moves into the foundry business, this announcement, when coupled with application in mobile market products, would certainly put Intel into a serious competitive position against Samsung and TSMC for Apple business, as has been speculated (see for example this discussion by EETimes).