Apple has finally announced its next-gen A14 SoC, succeeding the existing A13 chip. The primary highlight of the event was the new 5nm process that powers the A14. Technically speaking this isn’t an achievement on Apple’s side. Rather, it’s TSMC that’s really responsible for this feat, beating rivals Intel, Samsung, and Global Foundries in the race to build the smallest, most efficient transistors (for not at least).
Furthermore, Apple’s official figures all compared the A14 to the older A12 instead of the preceding A13 part which has been out for quite a while now. On the CPU side, Apple is sticking to the 2+4 configuration, with 2 hi-perf cores and 4 low-power cores. Apple claims an uplift of 40% compared to the A12 here but doesn’t really elaborate on this. Is it single-threaded or multi-threaded? I believe its the overall multi-threaded performance. Normalizing it against the A13’s gains over the A12 and we get a roughly 15-16% faster CPU. Not bad but considering the node shrink, it’s hard to figure out why we didn’t get a larger uplift.
As for the GPU, Apple promises a 30% gain in performance compared to the A12. This is primarily due to the new 4-core GPU. However, when we compare it against the A13, it comes to just 8%. Rather paltry, isn’t it?
The area where Apple actually seems to be making progress is its neural engine. The A14 has a 16-core neural engine with a rated performance of 11 TOPs, around 83% higher than the A13’s 6 TOPs.
While one can’t be sure behind the rather timid performance figures of the A14, it’s likely that Apple is focusing more on battery life and efficiency this generation. The 5nm EUV node from TSMC is already ahead of the curve in that space and paired with lower clocks, it should really hit the spot, even if the performance isn’t all that different.