The A15 chip integrates up to 15 billion transistors and is equipped with a new 6-core CPU, including two performance cores and four energy efficiency cores, which improves CPU performance by 50%.
What is the concept of 15 billion transistors, which is very close to the 16 billion transistors of the M1 chip, which is a new chip on the desktop.
The CPU is still the same, consisting of two performance cores and four efficiency cores, and the biggest changes should be GPU modules and AI chips, which is rare for Apple to realize the design of different configurations with the same core in the iPhone.
At this conference, A15 was also divided into three versions, one is to remove a GPU core, the other is to reduce the CPU frequency, and there is a full blood version. The latest iPad mini 6 just comes with a neutered version.
The A15 chip in the iPhone 13 and iPhone 13 mini is a 4-core GPU, while the two Pro models and the iPad mini 6 are a 5-core GPU full blood version.
？
A15 chip performance improvement over A14
It is understood that the A15 bionic chip in the CPU part of the improvement is not obvious compared with A14, at the press conference, Apple often uses A12 and A15 to highlight the obvious generation difference. However, from the Geekbench score, the CPU single-core performance of the A15 is 10% higher than that of the A14, and the multi-core performance is 18%, while the GPU improvement is more obvious, about 55%.
Of course, peer products and A15 chip still have a certain performance gap, Apple even directly said at the press conference that compared with \"competitive products\" CPU performance is 50% stronger, GPU performance is 30% stronger.
CPU performance is almost unchanged compared to the previous generation, the performance improvement of the GPU part is closely related to the increase of the number of GPU cores, and the actual performance of the A15 image performance of the new 4-core version may be weaker than another grade.
？
AI is becoming more and more important to the iPhone
In the A15 chip, the brightest performance improvement is undoubtedly the AI neural engine, although it is still the same 16-core architecture as the A14, but the computing power has increased from 110,000 operations per second to 15.8 trillion operations per second, a full increase of more than 40%.
The improvement of AI computing power does not bring the same effect as CPU and GPU, and the performance improvement of the latter two is basically a vertical increase in fluency, while AI computing power is a horizontal functional improvement.
Face unlock, through efficient face recognition algorithm, photo beauty function. When taking photos, AI can intelligently analyze the user's age, skin color, body type and other characteristics, and accurately improve the image quality through AI algorithms such as face detection, key point detection, and scene recognition, and automatically beautify the face, so as to make the people in the photos more beautiful. These technologies are not without the support of A15 chip AI computing power.
Under the trend of decreasing the impact of performance on smartphone experience, AI's transformation and upgrading of mobile phone functions will only increase.
？
A14 chip performance improvement over A13
The A14 processor uses TSMC's 5nm process and has 11.8 billion transistors. The A14's CPU core architecture is still designed with 2 large cores and 4 small cores.
According to the data previously released by TSMC, the 5nm process relative to its previous 7nm process, the logic density is increased by 80%, the performance is improved by 15% at the same power consumption, and the power consumption can be reduced by 35% at the same performance.
The A13 offers a 20% improvement in both CPU and GPU performance over the A12. Therefore, if the processor performance metrics of the two generations are the same, the A14's CUP performance is only about 16% faster than the A13, and the GPU performance is only about 8.3% better than the A13.
If you exclude the addition brought by the 5nm process itself, compared to A13, the A14's performance improvement in the CPU and GPU core itself may be very small.
However, in the AI computing core, the A14 has a very large improvement compared to the A13, from the previous 8-core NPU to 16-core NPU architecture.
According to data released by Apple, the AI computing power of A14 has also increased to 11.8 trillion times, and the machine learning speed has increased by 70%.