On November 10, Apple introduced three new Mac devices — MacBook Air, MacBook Pro and Mac Mini — powered by the Apple M1 system on a chip (SoC). While the inability of Intel chips to meet emerging requirements was the apparent trigger for Apple to have own chips, there were many other benefits and possibilities pushing the move.
Just like iOS and Android, Apple’s macOS has a different ecosystem than Windows. Although macOS has continued to improve its user interface, Apple’s service revenue from the PC segment can’t be compared with that from the iPhone. The slow improvement in PC hardware, which is dominated by Intel, is among the key factors affecting macOS’ progress. The gap between the two operating systems is also getting smaller as Windows is catching up. On the other hand, thanks to the healthy competition between the two major smartphone operating systems, smartphones have improved system architecture, such as memory management, power management, artificial intelligence functions, and user interface. They have integrated new chips, such as UWB, GPS and LIDAR, to meet the AI and IoT demand. The OS is undoubtedly the critical determinant of system performance. However, the most vital remains the brain — microprocessor.
Intel dominated the entire PC architecture for several decades. Many peripheral connections (such as wired and wireless connectivity), memory configurations, and integrated GPU and AI chips remained constrained by the Intel architecture. As a result, PCs and NB ODM/OEMs saw limited innovation. They just (or could only) followed Intel’s roadmap for x86 architecture advances. x86 has not been able to create many innovative applications to fulfill the emerging requirement for AI and IoT.
To overcome this, Apple announced at this year’s Worldwide Developers Conference (WWDC) that its entire Mac product line would switch from Intel chips to Apple Silicon within two years, and officially released the Apple M1 on November 10.
The M1 specifications are similar to those of the A14 Bionic. However, the M1 on a Mac device can deliver more computing power and applications than the A14 Bionic on an iPhone. Both are manufactured at TSMC’s 5nm and feature custom high-performance (Firestorm), energy-efficient (Icestorm), GPU and NPU. However, the M1 has two more high-performance cores than the A14 Bionic, as well as additional cache and peripheral circuitry.
The purpose of the first generation of Apple Silicon, M1, is to create a new experience. It aims to achieve the balance of performance and power consumption, instead of high computing. According to Apple, the M1 has 16 billion transistors, about 35% more than the A14 Bionic. Counterpoint estimates that the M1 die size is approximately 140-150mm2, much larger than that of the A14 Bionic.
Depending on the applications’ environment, power consumption, thermal efficiency and footprint are essential considerations in designing SoCs. Thanks to the advantages of the advanced process in TSMC in terms of power consumption and transistor density, the M1 performs better on Mac devices than on previous Intel-based MacBook. Unlike Intel CPUs that emphasize a maximum clock rate, the Apple M1 clock rate is dynamic. The M1 utilizes energy-efficient cores to achieve lower system power consumption and multitasking workload. However, in addition to the limitations inherited from the Apple AX SOC, the M1’s maximum clock rate cannot boost to 4.8GHz like Intel’s latest Tiger Lake CPU (10nm), which is more likely due to TSMC’s process limitations.
It is worth mentioning that the M1 embeds DRAM next to the SoC to form a unified memory architecture (UMA) to speed up data access through high-speed inter-connections. In contrast to the A14 Bionic, the DRAM is placed next to the SOC instead of being stacked. This idea has been mentioned in many documents, but its use in a PC is groundbreaking. As Michael Dell once said, “Ideas are commodities, execution is not.” Apple has managed to prove the benefits of UMA through actual product design and validation.
Below are the advantages of this packaging method:
- The embedded DRAM in the M1 can reduce memory modules’ footprint, leaving room for batteries or other peripheral circuits.
- The signal transmission speed between the SOC and DRAM in the same package is much faster than in the design where the chip is separated at different positions on the PCB. Of course, it becomes more difficult to manufacture such chips. But TSMC’s advanced packaging technology enables such a design.
- Unlike the legacy architecture in traditional PCs, the M1 embeds high-bandwidth and low-latency memory DRAM into a single pool within a custom package. Therefore, all cores can retrieve data in the memory simultaneously, and the system can dynamically arrange the precious memory resource. This dramatically improves system performance and reduces the usage of power-hungry DRAM, consequently extending the battery life.
As we mentioned in a report published in June, the benefits of migrating to the new architecture (Apple Silicon) include:
- Reducing reliance on a single supplier
- Differentiating products from competitors
- Gaining control of the ecosystem
- Developing apps for both mobile and computer using the same APIs and programming tools.
- Reducing cost
The usage of the M1 in Mac devices demonstrates these benefits. In terms of cost reduction, the M1’s production cost is lower than the Intel CPU’s purchase cost. Therefore, the overall price of the system will be lower than before.
Further possibilities with Apple Silicon include:
- Multi-cores: The next generation of Apple chips will have more high-performance cores to meet the requirements of multitasking and high-performance PC applications. However, this will inevitably be accompanied by an equivalent increase in embedded SRAM density, leading to a significant cost increase. Therefore, if the technology migration becomes slower, dual Apple Silicon coexistence may be a potential solution.
- Powerful NPU: The NPU allows the Mac devices to enhance AI inferring capabilities in image and voice, such as integrating virtual assistant (macOS Siri) to enhance the Mac’s strengths in the office, reducing the bandwidth required for remote video-conferencing and providing instant translation at the edge.
- Integrated discrete GPU chips: Thanks to its high-performance GPU core driven by TSMC 5nm, the M1 is more powerful than the mid-range standalone GPU graphics cards. However, it is still insufficient for content creators. Therefore, Counterpoint believes Apple Silicon will integrate AP and GPU into a single chip by using advanced package
(To be continued. This is Part I of a series)