Intel's Tri-Gate gamble: It's now or never
Deep dive into Chipzilla's last chance at the low end
Fast forward to 2011: Chipzilla under threat
The situation today is quite different. While Intel still squats in more server sockets than its AMD competition, down in the consumer space its hegemony is in question. The market-watchers at IDC, for example, predict that ARM will suck up 13 per cent of the PC marketplace by 2015, due mostly to Microsoft porting Windows 8 to ARM in 2012 or 2013.
AMD is a resurgent threat, as well, with its new Fusion line of CPU/GPU mashups that AMD insists upon calling APUs – accelerated processing units. Intel's vastly greater marketing and distribution resources will surely make it difficult for the "li'l microprocessor designer who could" to tear significant chunks of market share away from Chipzilla, but don't think they won't try – and, with ATI graphics chops on their APU silicon along with new Bobcat and Bulldozer cores, don't think they won't succeed to an Intel-annoying degree.
Down in the exploding tablet and smartphone space, Intel is, well ... to call their current position "suboptimal" would be kind. The company's pioneering low-power efforts – "Menlow" and "Moorestown" – were essentially failures, and the jury is still out on its most recent effort in this space, "Medfield".
ARM – with its IP being embodied in silicon by chip-bakers such as Samsung, Qualcomm, Nvidia, TI, and others – owns the tablet and smartphone market.
But what must be even more troubling for Intel is that ARM is moving up into the PC market – as evidenced by IDC's predictions noted above, and soon to be boosted by the consumer-level, PC-capable ARM Cortex-A15 design, scheduled to appear next year in chips from TI, Nvidia, and others.
Just to put an additional frisson of fear into Intel's corner offices came a rumor this Friday that Apple – Intel's BFF since the first Chipzillian Macs shipped in 2006 – is mulling a switch to ARM for its hella-popular laptops (at minimum) in the next two or three years.
Intel is finding itself threatened at the growing low-power end of the consumer spectrum in a way that it has never been before. (It's also facing challenges in the server space from such new developements as microservers and massively multicore processors from outfits like Tilera, but those threats are not as immediate.)
So when a company is threatened, what does it do? It falls back on its strengths. And what are Intel's strengths? Engineering smarts and cash. And it took both to bring the Tri-Gate design and manufacturing process into fruition.
You might also argue, of course, that another of Intel's strengths is the gargantuan installed base of Intel-architecture apps littering God's green earth – and you'd have a good argument, if you were talking about PCs. Tablets and smartphones, not so much – which is, of course, why Intel is busily porting Android's tablet-centric version 3.0 to IA.
But exactly how good is 22nm Tri-Gate, and will it be fast enough and low-power enough to move Intel into the range it must comfortably inhabit if it's going to compete with ARM?
The simple answer to that question is: "Dunno." After all, it will be a while before we get our hands on 22nm Tri-Gate microprocessors. All we have so far are presentations and promises. However, if what Intel revealed on Wednesday is even close to the truth, there a chance that the first procs to be built using the process, which will be code-named "Ivy Bridge", might be worthy competitors.
With the help of Intel senior fellow Mark Bohr, who provided a technical overview of Tri-Gate at the rollout – plus enduring one of the more embarrassing marketing-video starring roles we've seen in some time – let's take a bit of a deep dive into the transistor tech that Tri-Gate embodies.