Apple's (AAPL 0.50%) new A11 Bionic chip that'll be at the heart of every iPhone 8, iPhone 8 Plus, and iPhone X seems like a truly incredible piece of engineering. Apple claims solid improvements in performance and capabilities across the board, and early performance tests appear to validate Apple's claims.
The A11 Bionic is also believed to be manufactured using Taiwan Semiconductor's (TSM 0.27%) new 10-nanometer technology, which is believed to offer substantial performance, power, and area improvements over the 16-nanometer technology that's used to manufacture Apple's previous-generation A10 Fusion chip.
Image source: Apple.
Although Apple has clearly built a best-in-class mobile chip that should deliver a mobile-computing experience that its competitors probably won't be able to touch for a long while, there's one way in which Apple played it relatively safe with the A11 Bionic.
It's probably a tiny chip
Apple says that the A11 Bionic is made up of 4.3 billion transistors (think of transistors as the building blocks from which chips are crafted). That's a lot of transistors to be sure, but it's not a big increase from the 3.3 billion transistors found in the prior-generation A10 Fusion chip.
TSMC's 10-nanometer technology should deliver roughly a doubling in transistor density -- or, put another way, a halving of the area that an individual transistor takes up.
The transistor count of the A11 Bionic indicates that Apple used the more advanced 10-nanometer manufacturing technology not just to add more transistors but to try to significantly reduce chip area while doing so.
To get a sense of how small the A11 Bionic likely is, the Huawei Kirin 970 chip -- also manufactured on Taiwan Semiconductor's 10-nanometer technology -- includes 5.5 billion transistors in an area of 100 square millimeters. If we assume similar transistor densities for the A11 Bionic and the Kirin 970 (a reasonable assumption since they are chips targeting the same basic use cases and power envelopes), then the A11 Bionic should measure approximately 78 square millimeters.
This would make it significantly smaller than the A10X Fusion -- found in the latest iPad Pro tablets -- which TechInsights measures at about 96 square millimeters.
Why did Apple play it conservatively with the A11 Bionic chip?
You must remember that by the time Apple launched the A10 Fusion chip, the 16-nanometer technology that it was built on had already been in mass production for a full year. This meant that yield rates -- the percentage of built chip products that are actually usable -- were probably robust, giving Apple's chip team the freedom to build larger chips without worrying about cost and/or supply issues.
Moreover, since TSMC's 16-nanometer technology was heavily based on its 20-nanometer technology -- meaning that the yield learning that TSMC did in ramping up the 20-nanometer A8 into production positively impacted TSMC's 16-nanometer yield -- the A10 Fusion really had the perk of being manufactured on a technology that benefited from two major A-series chip ramps.
TSMC's 10-nanometer technology, on the other hand, is brand new. It features a significant area reduction from the 20-nanometer and 16-nanometer technologies, and I wouldn't be surprised if the yield rates of 10-nanometer chips this early in the production ramp-up were far from ideal.
People at Apple aren't stupid; they almost certainly knew well before the A11 Bionic went into production that they would be landing the chip on a brand-new technology that didn't have the benefits of being built substantially on previous-generation technology.
I suspect it is this knowledge, coupled with Apple's desire to not have A11 Bionic chip manufacturing bottleneck production of the iPhone 8, 8 Plus, and X, that led to Apple's chip teams building a relatively small chip for this iPhone cycle.
Remember that smaller chips, all else equal, tend to yield better than larger chips. And, of course, more smaller chips can be produced on a given silicon wafer, further lowering cost.
What to expect from the A12 and beyond
I believe that the next A-series chip, the A12, will be manufactured using TSMC's first-generation 7-nanometer technology and that the follow-on chip, the A13, will be manufactured on TSMC's second-generation 7-nanometer technology, known as 7nm+.
TSMC has indicated that its 7-nanometer technology will build upon its 10-nanometer technology, as well as provide both area and performance improvements over the 10-nanometer technology, so I wouldn't be surprised to see Apple get slightly more aggressive in terms of chip area with the A12 compared to the A11 Bionic.
For the A13, given that TSMC's 7-nanometer technology should be fairly mature by the time it goes into mass production, I'd expect Apple to be more liberal in its use of transistors and ultimately expect non-trivial chip size growth to accommodate a wide array of new features.
