This year, Apple (AAPL 1.45%) announced a trio of new iPhones, each of which is powered by Apple's custom-designed A11 Bionic chip. The A11 Bionic is an impressive chip, offering solid improvements in CPU and graphics performance over its predecessor as well as enhancements that help boost battery life, camera and video image quality, and more.
Image source: Apple.
It's a great engineering achievement on Apple's part.
Despite that, I suspect that, thanks to issues with the Taiwan Semiconductor Manufacturing Company (TSM +0.40%) 10nm technology that the chip is built on, Apple missed some of its performance targets for the chip.
Allow me to explain.
A smaller-than-typical performance boost
In the table below, I show Apple's claimed CPU and graphics performance increases over prior generation chips for each A-series processor generation going back to the A8 chip, which Apple introduced in the fall of 2014:
| Chip |
CPU Improvement (%) |
GPU Improvement (%) |
Manufacturing Tech. |
|---|---|---|---|
|
A8 |
25 |
50 |
20nm |
|
A9 |
70 |
90 |
14nm/16nm |
|
A10 |
40 |
50 |
16nm |
|
A11 |
25 |
30 |
10nm |
Source: Apple.
In terms of generational CPU improvement, the A11 Bionic represents the smallest gain since the introduction of the A8. Generational GPU improvement for the A11 Bionic is the smallest in four generations, coming in even worse than the A8's improvement.
What's even more interesting is that Apple delivered a much larger boost in both CPU and graphics performance when it went from the A9 to the A10, which were both manufactured using TSMC's 16nm technology, than it did when it moved from TSMC's 16nm technology to TSMC's 10nm technology.
It might be tempting to blame this on Apple's engineering rather than on the underlying chip manufacturing technology, or it might seem reasonable to argue that at some point generational improvements become harder to come by as performance levels reach new heights.
Though I can't rule out either for certain, I think TSMC's 10nm technology simply didn't deliver the kind of performance that Apple (and likely TSMC) had hoped for (something that industry analyst Daniel Matte believes to be the case), which ultimately forced Apple to run the A11 Bionic's CPU and GPUs at lower frequencies than it had hoped to.
TSMC 7nm should be a lot better
I think TSMC's 10nm technology is analogous to its 20nm technology used to build the A8 -- it delivered a big improvement in chip area scaling, but it didn't enable a significant improvement in frequency, which ultimately held back performance.
The good news is that I think TSMC's upcoming 7nm technology, which will almost certainly be used to manufacture most, if not all, of Apple's upcoming A12 processors, will be analogous to TSMC's 16nm technology -- a technology that doesn't deliver as aggressive an area reduction but enables a large performance boost.
Indeed, per a paper that TSMC presented describing its upcoming 7nm technology at the International Electron Devices Meeting (IEDM) last year, TSMC's 7nm technology "provides >3.3X routed gate density and 35%-40% speed gain or >65% power reduction" compared to the company's 16nm technology.
I think Apple's A12 will, enabled by TSMC's 7nm technology, deliver a much larger boost in both graphics and CPU performance compared to what the A11 Bionic did. This should help improve the user experience for next generation iPhones, particularly as the new phones may have Apple's fast ProMotion displays, which will require much more computing power than standard iPhone displays do.
