Apple’s A16 CPU is Quite Underwhelming for Developers

Year after year, Apple introduces a new A-series system-on-chip to go along with the newest iPhones. With the iPhone 14, the firm has taken a completely different direction. The new A16 CPU is exclusive to the “Pro” model iPhones, while the normal iPhone 14 models retain the A15 chip launched last year.

A16: How is it Different from A15?

At first appearance, the A16 seems to be architecturally identical to the A15. To operate machine learning and AI algorithms, there are two high-performance CPU cores and four high-efficiency cores, five GPU cores, and 16 Neural Engine cores. The same as the A15.

According to Apple, the CPU is made using a novel “4 nanometer” technology from TSMC, making it the first such processor in a smartphone. It should be noted, however, that TSMC’s “N4” technology is not a real 4nm process. While it is a more sophisticated technology than previous A series CPUs, it is not a true next-generation silicon fabrication method; for that, you will have to wait until next year for the 3nm process.

The transistor count has increased by a few percent to 16 billion, with the majority of the extra money likely going toward the new display engine, which manages the iPhone 14 Pro’s display down to 1Hz in always-on mode and can crank it up to 2,000 nits in bright sunlight, memory controller, and image signal processor.

The processor’s more general-purpose components seem to have altered just little. The high-performance CPU cores are code-named “Everest,” and they can clock up to 3.46 GHz, which is around 7% faster than the A15’s top of 3.24 GHz for its “Avalanche” cores.

The high-efficiency cores are code-named “Sawtooth” and have a clock speed of 2.02 GHz, which is about the same as the A15’s 2.01 GHz “Blizzard” cores. While these cores have a new name, the architectural changes seem to be minimal at best, since they do not provide performance beyond the predicted rise in clock speed. Hence do not expect a major difference in processing power from the A15, any heavy processing apps such as clone it will more or less work at the same efficiency.

The Neural Engine retains its 16 cores from the A15. Apple claims it can do up to 17 trillion operations per second, an 8% improvement over the A15’s 15.8 trillion. The move to LPDDR5 memory, which should deliver 50% greater memory bandwidth than the A15’s LPDDR4x memory, is perhaps the most important change. Apple did move to LPDDR5 in the M1 processor line, which is based on the A14 chip architecture — the only surprise here is that the firm waited this long to do it in iPhone-bound processors. There may be certain extremely particular instances when a job on the A15 is fully restricted by memory bandwidth, in which case the A16 should perform much better.

Overall, we have what looks to be a faster A15 with a new display engine and maybe an image signal processor.

Wrap Up

The A16 is definitely not a “binned” variant of the A15. This is a brand-new chip. However, there are no substantial architectural changes here, just small tweaks to increase maximum clock speed and power efficiency. This is a smaller improvement over last year’s model than we’re accustomed to seeing in Apple’s annual iPhone refresh.