Apple A14 Bionic is the new mobile SoC that the Cupertino firm has created for the iPhone 12. Apple has shown that without being a firm specialized in semiconductors, its designs not only compete with those of Qualcomm and Samsung, but also surpass them in performance generation to generation.
What’s inside this SoC that makes it so fast? ICmasters, a semiconductor reverse engineering and IP services company, has decided to find out by putting it under the focus of a transmission electron microscope (TEM) capable of magnifying an object up to a million times.
Apple A14 Bionic under the microscope
This chipset is the most advanced ever designed by Apple. Surely it has had to do that in addition to motorizing its new generation of mobiles is that it is the basis of the A14X, the main engine of the Apple Silicon platform with which the Cupertino giant wants to revolutionize the industry by replacing the Intel x86 hardware that He’s been using Macs on his personal computers since they replaced the Power PCs.
The first technical section to highlight comes from its 5 nanometer manufacturing process. It is the first SoC to achieve this level of integration in a production under the responsibility of the TSMC foundrie. Under an architecture licensed to ARM Holdings, the die includes many components, the main ones:
- A six-core CPU that can scale to around 3 GHz.
- A GPU with four graphics cores.
- Security chip for secure enclave.
- A digital signal processor (DSP).
- A 16-core neural processing unit.
- Other co-processors.
Viewed under the microscope, the size of the matrix is 88 square millimeters, tremendously small considering that it includes 11.8 billion transistors. However, the density metric does not correspond to that of TSMC. Instead of 171.3 million transistors per mm2, the ICmasters measured 134.09 million transistors per mm2.
This is a big difference, although each layout is different due to different logic and cache layout. Historically, Apple chips have reached 90% of the theoretical compute node density in their processors. This generation stands out for losing that brand to a great extent.
The Apple A14 Bionic has a 78% effective transistor density compared to the theoretical density. There is plenty of scope to include more transistors if as you suspect this matrix will be used in the A14X for Mac.
Another detail of this development comes from the smaller space occupied by the SRAM. A typical mobile SoC consisted so far of 60% logic, 30% SRAM and 10% analog components, they explain from TSMC. SRAM’s 3D stacked technology will help alleviate the density issue and make room for other components.
Regarding costs, the calculation made by TSMC indicates that the price per transistor would not have decreased in the transition to N5 from N7, as it would have happened from the previous nodes. It is clear that in technological manufacturing processes as advanced as 5 nanometers, the margin is much narrower in any parameter.
The microscope image shows the detail of the processing cores. Both ‘FireStorm’ high-performance together with its cache use 9.1 mm2, while the other four cores ‘IceStorm’ dedicated to light tasks and containment of consumption occupy 6.44 mm2. The GPU group and its four cores use approximately 11.65mm2. There is a unified cache that, they say, cannot be found under the microscope.
Interesting is this Apple A14 Bionic that confirms the ability of the Cupertino firm in the design of its chips and also that of TSMC as a producer. The great performance result in the iPhone 12 indicates that Apple can do very well when it brings the ARM architecture to the Macs.