The latest crop of Qualcomm(QCOM +0.62%)Snapdragon chips, from the relatively low-end Snapdragon 400 series to the highest-end Snapdragon 810, all use standard ARM (ARMH +0.00%) CPU cores. Moreover, many of these chips implement ARM big.LITTLE technology, pairing very high-performance cores for more performance intensive workloads with low-power cores for less intensive workloads.
Although Qualcomm is likely to drop big.LITTLE in its next-generation Snapdragon 820 (which features a custom-designed Qualcomm CPU core), will the company continue using big.LITTLE in its mid-range chips?
Will Qualcomm use its custom CPU core in the low-end and mid-range?
Over the last several years, Qualcomm has tended to use its own custom-designed cores in its high-end chips, while employing standard ARM CPU cores in the mid-range and low-end. My guess is Qualcomm's custom high-end CPU cores will probably take up more silicon real-estate -- and potentially offer better performance per watt -- than the highest-end off-the-shelf cores from ARM.
This would mean Qualcomm would use its homegrown CPU core in chips commanding a big premium and then ARM-designed CPUs in chips that cost less.
If that turns out to be the case, and given that all of Qualcomm's ARM-based competition in the mid-range will be using big.LITTLE in order to advertise "more cores," I would expect Qualcomm to keep using big.LITTLE in its mid-range parts.
Does this create a marketing problem for high-end Qualcomm parts?
If Qualcomm is out marketing six and eight-core processors in the mid-range, will it have a tough time selling what are likely to be quad-core parts at the high-end? At one point, I would have thought so, given how fiercely competitive the mobile market is and how focused people can sometimes get on specifications.
However, at the high-end of the mobile market today, the three major chip design teams are Qualcomm, Samsung, and Apple. Samsung uses big.LITTLE in its flagship Exynos chips, but the company has signaled that it, too, is designing a custom ARM core. Somehow, I doubt it will use big.LITTLE there.
Apple is still rocking two very powerful cores in its A-series processors, and I do not expect the company to sacrifice real-world performance and power efficiency for marketing. It also works in Apple's favor that it does not advertise core counts in its marketing messages to end users -- they do not expect a certain core count, just better performance each time.
So, if Samsung is probably going to move to a quad-core custom solution for its own Exynos chips, and if Apple is unlikely to move to quad-core solutions anytime soon, then at the high-end, Qualcomm should have no problem successfully fielding quad-core solutions.
Remember that Qualcomm is not really trying to sell chips to end users -- it is trying to sell them to phone vendors. As long as its chips can offer best-in-class power consumption and performance, then CPU core count, among all of the various elements of a complex mobile system-on-chip, is not going to make or break a design win.
