In a recent column, I argued that Intel's (NASDAQ:INTC) Atom mobile processor cores had fallen behind in performance relative to what ARM Holdings (NASDAQ:ARMH) offers to its licensees. Though the company continues to develop future iterations of its Atom processor line, I remain skeptical that the company is devoting sufficient resources to those processors in order to gain a leadership position.
That said, although the company seems to be treating its Atom processor cores as second-class citizens, Intel continues to invest heavily in improving the performance and power consumption of its mainstream "big" processor cores.
Given that Intel has tens of billions of dollars of business riding on the competitiveness of these cores, it's little wonder the company works hard to make sure they are truly best-in-class.
In light of this, I believe it's time for Intel to begin using its Core CPUs to attack the high end of the mobility market (smartphones and tablets) in earnest.
Isn't that what Core M is for?
Intel has been trying to push its Core architecture into low power segments with its Core M product line. The first iteration of Core M, which was based on the company's Broadwell architecture, was essentially the same exact silicon as the company's Ultrabook-oriented processors, but mounted onto a smaller package.
A reasonable start, but hardly enough to truly compete with the various ARM-based system-on-a-chip designs currently in the marketplace.
The second-generation Core M chip, which is based on the company's Skylake architecture, actually brought some specific enhancements to make it more suitable for "tablet-like" usages. Such enhancements included:
- An even smaller, more compact package compared to the Broadwell Core M processor
- An integrated image signal processor
- Integrated of low-power I/O typically found in tablet-focused processors into the chipset
- Integrated sensor hub
These are all excellent enhancements, and I suspect that for 2-in-1 devices as well as ultra-thin fan-less notebooks, the Skylake-based Core M is a great product. However, in order to really go head to head with the various ARM vendors in phones with Core M, it will need to make a number of changes in future products.
What would those changes look like?
The tricky thing about trying to compete in the market for high-end smartphone applications processors is that it's not enough to win by just, say, offering better CPU performance or graphics performance than the competition.
In order to "win," a company needs to offer a product with leadership across all of the major intellectual property blocks. Such blocks in a smartphone-oriented processor include:
- Image signal processor(s)
- Sensor hub
In order to be suitable for flagship smartphones, I believe Intel needs to do the following in its next-generation Core M processor:
- Integrate a leading-edge LTE modem.
- Integrate the company's latest image signal processors (i.e., from Intel's Atom IP roadmap).
- Stay on the cutting edge of video encode and decode capability (with the current Core M, Intel has good support for leading edge codecs, but it will be outclassed by premium ARM chips, such as the Qualcomm (NASDAQ:QCOM) Snapdragon 820, shortly).
- Integrate all of the functionality provided by the platform controller hub (currently a separate die on the same package as the main chip) into the main chip.
When might we see such a product?
It would be nice to see such a product launch in the second half of 2017, when the company introduces its Cannonlake architecture. However, it's important to keep in mind that such a product would need to be built on the company's system-on-a-chip manufacturing process rather than its traditional CPU-focused process.
Given that such processes typically lag the company's CPU-oriented processes by a number of quarters, this seems unlikely.
Additionally, Intel would need to be able to get a leading-edge modem implemented in its 10-nanometer manufacturing technology and integrated in with the rest of the applications processor. Given that the company isn't expected to launch its first 14-nanometer stand-alone modem until early 2017, such a timeline seems far too aggressive.
If Intel really executes properly, then we might see such a product launch in late 2018, when the company introduces its 10-nanometer "tock" product family (i.e., major design overhaul) known as Icelake.