Intel (INTC +0.77%) plans to launch the follow-on to its TSMC (TSM +0.70%)-built SoFIA LTE, which integrated a cellular baseband with an applications processor aimed at low-cost devices, in the second quarter of 2016, reports DigiTimes. The chip giant also reportedly plans to launch its mid-range-oriented SoFIA MID, an integrated modem and applications processor solution aimed at mid-range mobile devices, in August 2016.
DigiTimes says that SoFIA LTE 2 will be a quad core part, as Intel previously indicated at its 2014 investor meeting. What's interesting is that the site reports that SoFIA MID will actually be an eight core part -- in contrast to a quad core part, as Intel had previously indicated.
What does this tell us about these parts?
If DigiTimes is correct in saying that SoFIA LTE 2 will be a quad core part and SoFIA MID will be an eight core part, then I'd bet that SoFIA MID will have eight Airmont cores rather than eight Goldmont cores (if SoFIA MID had eight Goldmont cores, then this would render the higher-end Broxton chip -- with only four Goldmont cores -- pointless).
On some level, this makes sense; in moving from the 28-nanometer process used to build SoFIA LTE to the 14-nanometer process used to build SoFIA LTE 2/SoFIA MID, Intel can pack a lot more functionality into a given chip area footprint. If Intel isn't going to use four higher-performance Goldmont chips, then I suppose eight Airmont cores would make the chip more competitive in the market than four Airmont cores.
To be clear, I fully agree with Qualcomm's (QCOM 0.84%) Anand Chandrasekher that octa-core mobile applications processors are "dumb"; the performance benefits of eight cores in real world mobile applications will likely be zero.
However, it likely doesn't cost Intel that much more to stick in four more Airmont cores in terms of actual chip manufacturing cost, and it gives Intel the "octa-core" marketing point that both Qualcomm and MediaTek have in their respective mid-range offerings.
Here's the issue, though
I think SoFIA LTE 2 might have a chance of being competitive (although I would have really liked to see it in early 2016 rather than the second quarter of 2016), but if DigiTimes is right about SoFIA MID being an octa-core part, and if I'm right about Intel using the Airmont CPU core in said part, then the competitive situation for that product looks iffy.
Qualcomm has already announced that its mid-range Snapdragon 618/620 will feature 2 or 4 ARM (ARMH +0.00%) Cortex A72 cores in a big.LITTLE configuration with four Cortex A53 cores, respectively, for the second half of 2015.
If Qualcomm moves these parts to either TSMC's (TSM +0.70%) 16-nanometer process or Samsung's (NASDAQOTH: SSNLF) 14-nanometer process by mid-2016, then such parts have the potential to outclass SoFIA MID as far as CPU performance goes due to the A72's higher performance relative to Airmont.
There's more to it than CPU performance, but it's important
A mobile system-on-chip is made up of many different blocks, including the cellular baseband, CPU, graphics, image signal processor, connectivity, and more. That said, Qualcomm has generally led Intel in terms of the non-CPU functionality found in mobile chips, so if Intel falls behind in CPU performance, I wouldn't expect any miracles from the rest of the chip.
I look forward to an update in November
By the time Intel's November investor meeting rolls around, the company should be sampling both SoFIA LTE 2 and SoFIA MID to customers. By then, the company should have a good idea of what kind of customer interest and design traction it's seeing with these parts.
Intel doesn't break out its mobile business separately any longer, but I'd bet that at the investor meeting it will give some color as to how its mobile business is progressing. If SoFIA LTE 2 and/or SoFIA MID gain meaningful traction in mobile designs slated to launch in 2016, we'll be able to see it in the company's full year 2016 guidance.
