I think we now know enough to make a few educated, but speculative, suggestions for the future development of the screamingly fast Apple Silicon chips used inside Macs, iPhones, iPads, and every other Apple product.
5G will be on-chip
Apple’s M1 chip combines the processor, GPU, Neural engine, ISP (image signal processor), memory, secure enclave, I/O controller and media encoding engines all onto one 5-nanometer System-on-chip (SoC).
Apple is also engaged in building its own 5G processors for its devices. (It currently uses those from Qualcomm.) The company is expected to begin deployment of its own self-developed 5G modems in iPads next year.
As Apple’s 5G technologies are iterated upon, it feels appropriate to predict these functions will eventually be hosted on the SoC, enabling the company to build iterations of the processor across all its devices. This would include Macs. That’s going to make for a networked future for WFH enterprises.
Apple Silicon goes to 32-core
Apple’s M1 processor as used in Macs hosts eight cores, including both high-performance and high-efficiency chips. The four former cores are designed to run a single task as efficiently as possible and maximize performance and can handle multi-threaded workloads astonishingly well. The high-efficiency cores use very little power to run but can handle all your lightweight tasks.
Apple already claims these chips can deliver up to twice the performance of the very latest PC laptop chip.
The eight-core chips inside the M1 are already winning high praise across the industry for their phenomenal performance, but this is only the beginning. Apple intends putting variants of the M1 chip inside future pro Macs (MacBook Pro, iMac, Mac Pro), and its road map sees a 32-core chip deployed inside of its highest end Macs, potentially by 2022. The company is said to be testing 16- and 32-core processors, with 12-core variants mentioned.
With Apple expected to introduce its silicon inside higher-end Macs in 2021, it seems likely that the highest end Macs will be running 32-core chips by 2022 at the latest.
Chips get smaller
The truism is that the smaller the chip the higher the performance and the lower the power demand. That’s why it’s interesting Apple partner TSMC is investing deeply in 3-nanometer process technology.
Assuming it all goes according to plan, the company is expected to deliver the first 3nm chips in volume in 2022. It makes sense for Apple to take a small-c conservative approach to a move to smaller die sizes, so I’d anticipate the first such chips may appear in an iPhone first and tweaked for deployment across the rest of its product line across 12-18 months – though development problems could slow this, and it is possible the company is already far ahead of expectations on this, given recent comments from executives.
Any move to 3nm should deliver even greater performance – the move from the 14nm A10 Fusion to the 7nm A11 processors unleashed a 25% performance bump. It’s reasonable to expect something similar.
What’s less predictable is when Apple might migrate to 3nm for M-series chips, but given it seems on schedule to ship 3nm processors inside iPhones at around the same time as it finishes its M1 to Mac transition in 2022, it’s reasonable to think 3nm Macs with even more astonishing performance will be the next major iteration on Apple’s then product development road map.
Beyond the CPU
It’s vital to remember that Apple makes integrated SoCs for its devices. That means its development teams aren’t just focused on the CPU, but iteratively developing every other component on chip – including the GPU.
The GPU is integrated on Apple Silicon. It offers up to eight GPU cores on the M1 chip, and Apple claims it to be the highest performing integrated GPU on any personal computer. As Apple ramps up the available cores on its CPU, it seems a no brainer to predict similar enhancement for the GPU.
A China Times report claimed Apple to be developing a GPU architecture codenamed Lifuka for deployment in the iMac.
Now, if Mac CPUs are on the road to 32-core, what happens if its GPUs take a similar journey? Given the GPU performance we’re already getting from the M1 chip, Apple’s teams seem to be treading a path in which Macs will deliver graphics experiences (and machine learning and every other task that makes some use of the GPU) that far exceeds that of any PC. Which should come in handy as the company iterates in related sectors and technologies, such as sci-tech, video encoding, and AR.
This same logic should be applied across every component of the chip. It’s a logic of iterative improvement that builds on the advances Apple’s processors already hold.
So, we now have a road map for Apple processor designs that leads us into 2022 and beyond. We have 3nm chips, vast advance in GPU designs, and an energy and performance matrix that means the highest end Macs are likely to set the bar for PC performance, even as iPads become capable of handling almost any task you want.
Connected by 5G and coupled with the huge advances in on-device real time machine learning the company is deeply focused upon, it’s possible Apple will reach a point at which its chips become so sophisticated the only way to advance them using current technologies will be to put these brains inside different product families, such as Apple glasses and vehicle navigation systems.
All of which the company seems to be working on.