Then on the process node side of things TSMC started to ship its first 28nm products to customers at the beginning of 2011, and we know that from there it took about 1 year to get GPUs and 1.5 years to get a smartphone using that node. TSMC is currently claiming that they'll be shipping 20nm in Q1 of 2013, so we might be seeing it used in phones by Q2-Q3 of 2014 if the trend holds. Oh, and TSMC is currently estimating that they'll start shipping 16nm in the 2014-2015 time frame.
The manufacturing lead Intel enjoys over AMD is significant because of AMD's dependence on GF's poor and inconsistent execution. In the ultramobile space that lead is significantly diminished or goes away entirely. If you consider that ARM is shipping 28nm chips already while Intel's still on their 32nm node, it's actually reversed. That's crazy to think about, but that mainly has to do with how unprepared Intel are on the low end. Medfield looks like a very good performer, but used in a very limited selection of products and still unable to distinguish itself from the rest of the pack. I see that as a good thing because it spurs innovation and competition between all of the players, but for Intel, a company who has enjoyed a significant lead in x86 over their biggest rival for years now, it's a shocker.
Win8 isn't going change this, either. You can't blame the economic downturn when last year Q4's sales are actually expected to outpace this year's with Windows 8 and all of its new devices hitting the market. Coupled with the very poor reception of Win8 in the eyes of the public and the early adopters (a majority of Win8 users prefer Win7), it would be patently false to claim that it's just an economic downturn that's at the heart of this.
Intel's response to the ARM threat and tablets has been Ultrabooks. They've missed their target for sales by a considerable margin and year-to-year sales and profit margins are expected to decrease. Although Haswell looks great, it doesn't look to tackle the biggest issue: price. In fact, if rumors are true, it's going to make them even more expensive.
So does Intel have a fab advantage over the ARM space? Sure, but it's not yet utilized and won't be unless TSMC fumbles again or Samsung encounters serious delays. On top of this, GloFo has already taped out ARM SoCs at the 20nm node (matching TSMC) and looks to ramp up production in late 2013 and 2014. Whatever fab advantage and performance gap that exists between AMD and Intel all but disappears at the mobile end. In order for Intel to enjoy the same gap over ARM that they do over AMD requires that their fabs run at full capacity and they enjoy their high profit margins, but the opposite has held true.
What does all of this mean? We finally get to see some cutthroat competition, and I, for one, can't wait
