According to a recent note published by Semiwatch's Robert Maire (via Barron's), semiconductor equipment vendors could be set to experience a sharp uptick of orders due to a 14/16-nanometer FinFET ramp up during the first quarter of 2015 in support of Apple's (AAPL -1.92%) A9 processor. That's certainly interesting, but in the note, Maire is quoted as having said the following:
If we were Apple we would be working on a back up plan if Samsung/TSMC can't get it [yields] up. Apple can't afford to miss the A9 in the fall of 2015. In fact Samsung executives have even bragged that they, not TSMC, will be producing the A9. This is a pretty bold prediction given the rumors that their yields still are stuck below 10% for 16/14NM FinFET. I wouldn't be so sure.
Now, while I'd argue that "stuck below 10%" is really more like "stuck between 10-20%," the interesting point of discussion here is what Apple could potentially have in place as Plan B.
Let's be clear: there's always a Plan B
For a company like Apple, whose two biggest revenue and profit streams depend on the supply of leading edge applications processors in large volumes, not having a Plan B would be reckless. So I'm confident that if both Samsung and Taiwan Semiconductor completely drop the ball on 14/16-nanometer manufacturing, there is a Plan B in place.
Then the question is -- what could this Plan B possibly consist of?
The obvious answer: a 20-nanometer A9 processor
SemiWiki's Daniel Nenni wrote back in August that he believed a 20-nanometer A9 processor seemed like a good, low-risk option. I wrote up my views on that possibility here and concluded that Apple would likely be able to deliver improvements over this year's A8 processor, but they would be relatively modest compared to what could be done on 14/16-nanometer FinFET.
Now, this would have some pretty interesting implications, if true.
Understanding the advancements that Apple needs to make each generation
The key parts of a mobile system-on-chip include the following:
- CPU -- controls how fast general applications run
- Graphics -- controls how fast games and other graphics-intensive apps run
- Imaging -- responsible for much of the "brains" behind the camera(s)
- Video decoder/encoder -- control the performance, features, and quality of video playback and capture
- Memory controller -- manages how data is passed between system memory and the chip
Now, the architectures of each of these blocks can be modified and improved independent of the underlying manufacturing process, but better underlying manufacturing technology increases overall power/performance.
So, what Apple could conceivably do for a 20-nanometer A9 is implement all of the architectural improvements that the 14/16-nanometer A9 would have had but then lower the clock speeds for some of the very high performance and power blocks such as the graphics and CPU. This would mean lower performance than the 14/16-nanometer A9 but still an improvement over the A8.
What is the downside to this?
One downside to this strategy is that it would take time and engineering effort to build a 20-nanometer variant of the chip as well as the 14/16-nanometer variant. Apple can certainly afford to do this, but it does seem rather wasteful.
Another downside is that the chip that debuts in the iPhone 6s/6s Plus will be the chip that powers Apple's lineup from late 2015 to late 2016. It seems likely that the Android competition, either through their own chips or through Qualcomm's Snapdragon processors, will launch phones equipped with 14/16-nanometer processors. This would put Apple at a significant performance/power disadvantage relative to its Android peers if it were stuck with the 20-nanometer node.
