Apple's latest iPhone, the iPhone 6s, in a new rose gold finish. Image credit: Apple.
In a recent research note from Cowen's Timothy Arcuri (via Barron's), the analyst says Apple (AAPL 1.19%) is "brewing evidence of three cores for [the Apple] A10" processor that will power next year's iPhone 7. He notes that such a move, which would mean an increased die size, would lead to increased demand for wafers and ultimately for semiconductor equipment vendors next year.
Although I concur with Arcuri that the A10 will almost certainly see die-size growth relative to this year's A9 (and the A10X may see growth relative to the A9X), I don't think Apple will add an additional CPU core. Here's why.
"More cores" is generally not the best way to go
The vast majority of mobile applications will benefit from improved per-core performance than from the addition of more cores. Many, if not most, mobile applications don't really know what to do with more than one or maybe two CPU cores, but every application can benefit from a core that can process a single application thread at a faster clip.
There is only so much power that a mobile CPU will be allowed to consume for an extended period of time. With a greater number of cores in the chip, the less power each core can be allowed to consume when all of those cores are lit up, lest battery life or thermals suffer.
Of course, many modern processors try to be smart about how power is allocated between the various cores; in a workload where all of the cores are lit up, the power may be distributed among the cores differently than it would be in a workload that really only pegs one or two cores.
However, if the CPU designer (in this case Apple) notices the vast majority of workloads benefit from allocating that fixed power budget to just one or two cores rather than across three, four, or even more, then it probably makes sense to include just two cores on the die.
Why aren't more cores better?
One question the reader might have is, "If the CPU can just dynamically allocate power as needed to the various cores, why not include more?"
Although there are many good technical reasons for this, there is an even more compelling economic reason: cost.
Even a rich and powerful company like Apple is subject to economic constraints. In order for Apple to be able to hit its desired per-unit profits, it needs to keep the manufacturing costs of its components in check. Although the cost of any one component that goes into a device like the iPhone is probably not terribly high, going over budget across many components could very well be "margin death by a thousand cuts."
So, we need to work under the assumption that, for each generation, Apple has a fixed amount of silicon real estate to work with. Apple's designers have the luxury of having more silicon real estate to work with than most other chip designers given the very high prices the company commands for its iPhones, but there are limits.
At any rate, Apple's CPU cores tend to be very powerful and, relative to cores from other vendors, quite large. According to a report written by Paul Boldt, the dual-core CPU complex inside of the A9 takes up a whopping 13 square millimeters, or around 14% of the total A9 chip area.
Adding another core would actually add a nontrivial amount of area, which would mean Apple would need to either incur increased production costs or it would have to skimp on other areas to keep the die size/cost in-check.
Given the minimal benefits associated with adding a third core, that previous die area is better spent beefing up other sections of the chip such as the graphics block (graphics is a special kind of workload in which "more cores," assuming the same type of cores, is better for performance).
Apple also gave us a pretty huge hint with the A9X
What's interesting is that even though last year's A8X chip found inside the iPad Air 2 employed a triple-core CPU design, the A9X chip found inside of the iPad Pro this year goes back to a dual-core design. Each core in the A9X is much faster than the cores found in the A8X, so much so that the A9X actually delivers better multicore performance than the A8X.
In light of the fact that Apple chose to use two higher-speed versions of the cores found in the A9 for the A9X -- which goes into a very large tablet that can handle much more heat from the chip than the thin and small iPhones can -- why would Apple suddenly add a third core to the iPhone 7-bound A10?
It doesn't make any sense. I believe that with the A10, Apple will make tweaks to the CPU cores themselves to deliver more "oomph" per clock and, thanks to a 16-nanometer process that's more mature, deliver higher clocks to boot. It will, however, most likely be a dual-core chip, not a triple-core one.
