Back in early 2014, Intel promised that the upcoming low-power implementation of Broadwell (Broadwell Y) would revolutionize what OEMs thought was possible with Intel x86 hardware. The new Core M platform would cut SoC power dramatically across the board, feature improved hardware integration, and allow for fanless “big core” x86 systems for the first time.
Unfortunately, history hasn’t quite played out that way. While Core M definitely found early success in some 2-in-1’s and ultrabooks, mainstream hardware (including newer 2-in-1 devices) has generally stuck to the Core i3 / i5 / i7 families. To put this in perspective, NewEgg shows an estimated 328 new 2-in-1 designs based on the Core i3 – i7 and just 42 systems built around Core M, M3, M5, or M7. While Core M isn’t absent from product line-ups, it’s certainly not as common. Digital Trends published a recent editorial arguing that “Nobody wants Intel’s Core M processor and Computex proves it.” While we’re not sure the situation is quite that stark, it’s true that Core M hasn’t become the mainstream platform of choice for modern devices, but we don’t think the problem is solely related to price or branding, which is where DT puts much of the blame.
Intel designed Core M to give OEMs more options than they’d previously had for building and differentiating high-end ultrabooks. Unfortunately, in many cases, OEMs chose to push for the thinnest form factors possible while simultaneously increasing screen resolutions. Pixels cost power, and if you have to run your monitor at 150-200% of default font scaling to read text in the first place, you’re literally throwing away battery power lighting up pixels you can’t see if run at their default resolutions. High-end displays also generate more heat, which means there’s less available headroom for the SoC itself. Since Intel chips still don’t ship with an integrated WiFi or cellular radio, they also wind up consuming more power on that front compared to ARM SoCs, which often have such components integrated on-die.
Intel has never confirmed if it builds Core M on the low-power variant of its 14nm process, but it’s hinted in the past that it does (and some of its PR literature refers to the 14nm process as distinct from its ordinary 14nm TriGate process). If true, it seems unlikely that the company would simply cancel the product family, but it’s also not clear where Core M will go from here.
Digital Trend’s battery tests revealed that while Core M was far more efficient than other chips at idle, it lost ground when compared against them in actual workloads. The website compared the Samsung Ativ Book 9, the Zenbook UX305, and the Lenovo T450s. The Ativ Book 9 uses an older first-generation Core M processor based on Broadwell while the Zenbook UX305 has a newer Core m3 6Y30 based on Skylake. Note that all efficiency figures have been calculated against the Lenovo T450s, a Core i5-5300U processor. The dramatic difference in these idle results for two Core M chips probably reflects subtle differences in UEFI tuning rather than fundamental differences between chip architectures, and they largely vanish in the later tests.
These real-world results suggest that while Intel succeeded in delivering a lower power part in absolute terms, it didn’t have as much luck building a chip with dramatically improved efficiency in conventional workloads. We’re not sure how much price would have impacted this, since Core M 2-in-1’s tended to be expensive, high-end products (cheaper price points, like the Surface 3, tend to use Atom processors, and the suggested list price on an Atom x7-Z8700 is just $37).
Ultimately, Digital Trends is right to think OEMs are moving away from the chip because it doesn’t offer enough appeal to justify its overall positioning — we’re just not sure cost is the principle reason. We’d argue that OEMs relentless focus on thinness, useless high resolution displays (relative to the size of the panels the notebooks / 2-in-1’s used) and stuffing smaller batteries into the resultant chassis played a part in limiting Core M performance and battery life. It’s no surprise, therefore, that many of these systems didn’t compare well against Core i3/i5/i7 platforms that were only slightly thicker.
If Intel does decide to kill the Core M family, it’ll presumably only do so when it moves to 10nm with Cannonlake. Since the 14nm process is already built and Core M implemented on it, there’s little to be gained by killing 14nm products while those lines are still cutting edge. And Apple has turned its Macbook into a Core M platform, which means Intel will crank these chips out for the foreseeable future.