Intel’s 14nm puzzle: As Skylake details leak, everybody asks – is the chip coming in 2015 or not?

Intel’s 14nm puzzle: As Skylake details leak, everybody asks – is the chip coming in 2015 or not?

Last week, we covered news of a new leaked Intel document that claimed to show the company’s 14nm Broadwell chips pushing back well into 2015. While new ultra-low power silicon will be available by the end of the year, this paper predicted that 14nm Broadwell-H processors — the more mainstream silicon, in other words — won’t ship until Q2 2015. Given this, we predicted that Broadwell’s successor, Skylake, would slip as well, possibly into early 2016. Now fresh documents show the opposite, with Skylake still coming next year.

As some of you have pointed out, the 2015 projection matches what Intel has said publicly, and what leaked roadmaps appear to show. One such roadmap is shown below:

This roadmap is already inaccurate — the i7-4790K launched in Q2, not Q3 — but it does show a combined Broadwell / Skylake product launch.

As for the new chipset roadmap, VR-Zone’s leaked documentsshow the Z170 chipsets arriving next year. These are mostly incremental updates to existing silicon, but the Z170 will supposedly offer support for Intel’s Super Speed Inter-Chip (SSIC) USB 3.0implementation. This is a new USB protocol that’s designed to allow for USB to be implemented as an interconnect between chips with minimal power consumption. It’s not clear how much of a benefit this will offer desktop users — SSIC is mainly aimed at the low-power mobile market, where it can cut power consumption by 80% in some cases.

Without any enormous changes on the chipset side, the question of which platform people buy into will be determined almost entirely by the CPU — and here’s where we split into two camps.

The argument for delaying Skylake holds that whatever problems Intel has had with Broadwellwill have inevitably impacted the launch of its entire 14nm product line. It was never clear that Broadwell would come to desktops — had Intel kept its entire launch schedule, Broadwell would’ve debuted earlier this year, the Devil’s Canyon Haswell refreshwould’ve kept the desktop side going, and Skylake would’ve dropped in next year. If Intel can’t launch Broadwell properly, it probably can’t launch Broadwell’s follow-up on time, either.

This argument makes sense if Intel’s goal is to keep the cadence of its tick-tock scheme, if the Broadwell delay hurt Skylake validation, or if the company simply sees no particular reason to push out a new architecture given the dearth of competition in its core markets.

But, as others have argued, maybe Intel does want to return to its tick-tock cadence. The only way to make that happen (and glean the associated PR benefits) is to push out Skylake quickly. I agree that this is technically possible — without information on why Intel’s 14nm is delayed, we can’t really speculate on the Skylake impact. If, for example, Broadwell was delayed by problems with scaling its fully integrated voltage regulator (FIVR), and Skylake ditches that design (as is rumored), than it’s possible that Intel can bring Broadwell to market and follow it quickly with a new microarchitecture. None of this is necessarily unprecedented — when Intel first launched Nehalem, the chip was desktop-only for nearly a year until Clarksfield launched in 2009.

Haswell (left) and the mobile Broadwell package

While it’s not unusual for Intel to run a bifurcated strategy for its products, the company has rarely launched two high profile new series so close to each other. It implies that Intel will either take a timing hit later down the road to reunify its strategy on 10nm (by skipping Skylake mobile) or that it will continue a staggered strategy of shifting to new architectures first on desktop, followed by laptop. One could call this a triplet instead of a tick-tock: new process (mobile) / new architecture (desktop) / new process + architecture (mobile).

Initially, I dismissed the idea that Skylake could launch directly after Broadwell, but after reconsidering the topic I’ll acknowledge that there’s no intrinsic reason Intel couldn’t do this. I still think it less than ideal — Intel has to choose if maintaining its cadence is more important than the costs it would face ramping up two different products in such a short time.  I think there are pieces to the puzzle still missing on this one — Intel could plausibly go either direction, but one choice pushes back product introductions while the other would incur additional expense and product line complexity.