Intel 3-D Transistor Plus Apple Equals Mobility Dynamite
Alone, the new transistor is a big leap forward. But the real advance could come on the partnership side.
Intel revealed what it promised would be "its most significant technology announcement of the year", namely the 3-D, or tri-gate, transistor being used in its latest 22-nm fabrication process. While it's easy to discount such vendor hyperbole as so much hot air, as a former semiconductor process engineer, I'm convinced this announcement warrants the hype as it represents a substantial technological leap beyond what competitors have anywhere near production--and that goes double if the company plays its cards right and explores an unprecedented partnership.
The 22-nm process represents the latest in Intel's tick-tock strategy for alternating major architectural changes to its x86 line with structural changes to the manufacturing process in successive product generations. The most recent "tick" was this winter's release of the Sandy Bridge CPUs, while today's "tock" will be manifested in the Ivy Bridge line late this year. Besides being an engineering and manufacturing tour de force, the release of a new architecture and process within one year shows that Intel's development metronome is speeding up. While the shrunken tri-gate process will obviously improve performance (faster clocks, bigger caches), it also should bootstrap Intel's efforts to break into the fast-growing, and heretofore impenetrable, mobile processor market.
And that's where things could get interesting.
Grasping the relevance to smartphones and tablets, and the reason process shrinks aren't always a benefit for mobile devices, requires understanding a bit of semiconductor physics. Bear with me here! Smaller transistors allow running circuits at lower voltage without sacrificing speed--in fact, the net effect of higher interconnect resistance (due to smaller wires) and capacitance (due to thinner insulators between wires) is more than overcome by faster-switching smaller transistors. Yet the smaller the transistor, the less it acts like a perfect switch--it's never entirely off, instead leaking current like a wall-wart power brick. Intel's three-dimensional transistor greatly mitigates this effect by applying electrical control to three sides of the transistor "pipe" instead of just one. To see an effective demonstration of this phenomenon, check out this Intel video, and to dive into the details (warning, technical hard hat area), look through its papers on this site.
The conventional wisdom is that by quickly migrating the tri-gate process to its Atom line, Intel can transform Atom from an underpowered cousin to its mainstream Core processors, only suitable for netbooks and set tops, to a smartphone and tablet powerhouse. Indeed, this is the likely desired scenario, since Intel needs an entree into the fastest-growing processor segment. But what if Intel decided to swing for the fences, swallow some pride, and team up with Apple to fabricate the A-series processors used on the iPhone and iPad? As a fabless developer, Apple designs these chips but relies on foundries to build them; it's currently using Samsung--and we know how well that relationship is going. Intel has never "leased" its formidable process technology to others as it represents a significant competitive advantage, but in the mobile market, the company is currently a non-entity. Yet teaming Intel's semiconductor prowess with Apple's market-leading design could lead to a hegemonic duopoly that utterly marginalizes others in the mobile device market. The question is: Will the egos at these two industry giants allow them to cooperate, or will they eventually collide?
Server Market SplitsvilleJust because the server market's in the doldrums doesn't mean innovation has ceased. Far from it -- server technology is enjoying the biggest renaissance since the dawn of x86 systems. But the primary driver is now service providers, not enterprises.