Intel’s SoFIA May Be Better Than Expected

One of the biggest disappointments in Intel's (INTC -0.38%) mobile road map has been the lack of an integrated connectivity/cellular with an applications processor for lower-end smartphones. At Intel's 2012 investor meeting, the company promised an integrated part built on its 22-nanometer FinFET process. The part never materialized. Now, at the 2013 investor meeting, the company announced plans for an integrated low-end part. While it's not quite clear what some may have expected or hoped for, it could end up being much more compelling than initially expected.

SoFIA: initially underwhelming...
At its investor meeting, Intel announced what appears to be a panicked reaction to a need to participate in the low-end/value segments of the smartphone market, and area that Intel believes will comprise roughly 50% of the smartphone market. After all, there's no way that the company can begin to participate in those markets without a highly integrated solution. Remember, these are smartphones that will be selling for roughly $50-$200, so the more that's integrated onto the primary system-on-chip, the better.

The chip family that Intel announced to target this space and that will be used to go head-to-head with Qualcomm, Broadcom, and MediaTek. What seemed initially off-putting about this particular chip was that it is slated to be built at TSMC (TSM 2.84%) on its 28-nanometer process. The reason is that Intel's modem/connectivity IP hasn't yet been ported to Intel's own internal processes -- an artifact of the fact that Infineon Wireless, Intel's modem team, had been working on these designs to be built at TSMC before it was acquired. So, without Intel's vaunted process advantage, what advantage could the company possibly have in this space?

It could come down to single-threaded performance
It seems that SoFIA -- both the 3G and the LTE versions -- will sport a port of the company's Silvermont processor core initially designed and optimized for the Intel's 22-nanometer FinFET process. Moving to 28-nanometer planar will lead to a performance reduction, probably manifested in the form of lower clock speeds within the same power envelope barring any major micro-architectural changes. But this isn't all bad. Why?

Well, in the low end of the smartphone market, other players such as Qualcomm, Broadcom, and MediaTek will likely stick to Taiwan Semiconductor's 28-nanometer process simply for reasons of cost -- 20-nanometer and 16 FinFET will be more expensive per transistor, so more suitable for higher-end, higher-ASP products. This means that as far as manufacturing at the low end goes, Intel will be on a par with its peers.

 While Intel's competitors will be pursuing four wimpy cores -- think ARM's Cortex A53 -- Intel will probably fight with two fast Silvermont processor cores. Intel will be at a marketing disadvantage, but in terms of the single-threaded performance that makes up the vast majority of mobile applications -- and potentially multi-threaded performance -- Intel's two higher-performance cores should have a real advantage on a technical front.

The 14-nanometer SoFIA would be a game changer
While Intel stated that its initial SoFIA would be built at TSMC, it plans to move SoFIA to its 14-nanometer process in 2015. This leads to a gross margin improvement; cost per transistor will decline, and it will not have to pay TSMC's foundry margin. It will also lead to dramatic performance and power improvement by virtue of significantly higher-performing transistors.

In short, Intel gains a pretty significant leg up on its competitors. Further, in the lower end of the market, where SoFIA is targeted, Intel's competitors may not be able to afford to move to FinFETs in the late 2015/early 2016 time frame. Intel's 14-nanometer node should have significant performance and density advantages over TSMC's 16-nanometer.

Foolish bottom line
With Intel on the same 28-nanometer playing field at the low end of the smartphone/tablet space as its competitors through most of 2015, it may have an architectural advantage, at least on the CPU side of things. However, by the end of 2015/early 2016, Intel will bring to bear its full manufacturing prowess, even in the low-end, high-volume parts of the smartphone market. While Intel needs to execute, the plan looks incredibly compelling to this long-term investor.