In 2023, after Apple launched its flagship iPhone 15, it quickly launched its second autumn launch event, with the theme of the Mac computer and the protagonist being the computer chip M3.
At a short 30 minute press conference on October 31st, Apple's M3 series of PC chips took the lead and launched three models: M3, M3 Pro, and M3 Max. Moreover, all three new products adopt 3-nanometer process technology and use a brand new GPU architecture, known as the "new king of Apple's personal computer chips".
The M3 series chips will be installed in Apple's new laptop MacBook Pro and desktop iMac. Among them, the starting prices for the new 14 inch and 16 inch MacBook Pro in China are 12999 yuan and 19999 yuan respectively; The new iMac will be launched next week, with a starting price of 10999 yuan in China.
Since Apple first announced its self-developed computer chip M1 in 2020, its chip family has become more systematic and powerful. It is not only a consumer electronics terminal enterprise, but also has the color of a semiconductor enterprise.
On the other hand, throughout the PC industry, the competition on the chip side is becoming increasingly fierce. Not only is Apple advancing on the M-series, but Qualcomm and Nvidia have already laid out PC CPUs. It is worth noting that the ARM architecture chips supported by these manufacturers are competing head-on with the x86 architecture, entering Intel's hinterland.
Guo Junli, Research Director of IDC Asia Pacific, told 21st Century Economic Report reporters: In the past, Microsoft and Intel have built a solid 'Wintel Alliance'. After decades of iteration, x86 has significant performance advantages, strongest compatibility, and supports the most applications. Therefore, Intel has always held a dominant position in the PC CPU, and AMD, as a follower, is constantly seizing the market. As more companies such as Apple, Qualcomm, NVIDIA, Samsung, etc. layout their ARM architecture PC chips, we believe that this trend will Intensifying competition, the CUP of ARM architecture will erode the market share of x86 architecture
Evolution of the M-series
Although this event was called the "shortest event in Apple's history," the M3 chip offensive was fierce, with performance progressing layer by layer from M3 to M3 Max.
Firstly, from the perspective of transistors, architecture, and performance. According to Apple's introduction, M3 has 25 billion transistors, 5 billion more than M2, and adopts the next generation architecture of a 10 core GPU, which can make graphics performance 65% faster than M1. At the same time, it has an 8-core CPU, which makes CPU performance 35% faster than M1. For ordinary users, the performance of M3 is already sufficient and there may even be an excess.
But Apple is still pursuing ultimate performance. The M3 Pro has 37 billion transistors, and the processing performance of an 18 core GPU is 40% faster than that of the M1 Pro. The single threaded performance of a 12 core CPU design is improved by up to 30% compared to the M1 Pro. In practical applications, MacBook Pro equipped with M3 Pro can perform a lot of panoramic photo processing and other operations in Photoshop more quickly. For users in some professional fields, high-performance can meet their needs.
Looking at M3 Max again, its transistor count has further doubled on the basis of M3 Pro, reaching 92 billion, which also means a significant improvement in performance. The 40 core GPU enables image processing speed to be 50% faster than the M1 Max and supports up to 128GB of unified memory, allowing AI developers to use large Transformer models with billions of parameters. The 16 core CPU enables performance to be 80% faster than the M1 Max. Obviously, this high-performance chip is designed for scenarios with extremely high computational power requirements, including large-scale video processing and generative AI applications.
From the data, the M-series performance monster continues to soar, especially worth mentioning the next generation GPU architecture in the new chip. It is reported that, unlike traditional GPU architectures, the new GPU has dynamic caching function, which can optimize the allocation and use of local memory, and run professional apps and games more efficiently. And for the first time, the M3 series of chips have adopted hardware accelerated ray tracing technology, which has also expanded from mobile phones to PC devices, enabling faster rendering and modeling, such as better shadows and reflections in games.
Guo Junli analyzed to reporters that due to the limited upgrade range of the M2 series chips, it did not fully stimulate consumers' enthusiasm for upgrading, resulting in a significant decrease in MacBook shipments. The launch of the M3 can help Apple improve this trend. Secondly, Apple's comprehensive upgrade of GPUs is in line with the current wave of artificial intelligence, and has the industry's pioneering dynamic caching function, which is quite friendly to visual developers such as games, revealing Apple's ambition to expand the gaming ecosystem. At the same time, the built-in neural engine of M3, M3 Pro, and M3 Max can accelerate machine learning models, echoing the trend of large model training.
Whether in GPUs or CPUs, Apple has been self developing for a long time, and Mac computers have completed the transition from Intel chips to their own M chips, reshaping the market landscape in PC terminals and chips.
From a PC perspective, industry insiders point out that from M1, M2, and now M3, Apple has solidified its computing power on the computer side in the iteration of the third generation. In the market dominated by Windows and x86, the combination of MacOS and Arm has formed its own moat. Although Apple Mac computers do not have a high proportion in the PC category, they have always focused on the high-end segment market, more like focusing on high profits rather than market share.
At the same time, Apple is selling products through an ecosystem. For example, as the iPad gradually introduces the M-chip, the interaction between tablets and PCs will be further upgraded. For example, Apple's entire series of phones, tablets, and PCs have already used the Arm architecture, making it easier to further collaborate on software and hardware. However, the Android and Windows camps are difficult to connect across the entire series.
PC Chip Competition
On the other hand, Apple has become a representative of PC chip competition, and its self-developed M-series chips are constantly crushing their competitors with every release. This year's market is even more lively, with Qualcomm and Nvidia accelerating their entry into the PC CPU field.
Just last week, Qualcomm launched the Snapdragon X Elite platform, a PC processor that focuses on AI, and X Elite was not born out of nowhere. In 2021, Qualcomm acquired chip design company Nuvia and subsequently launched the Oryon CPU based on the 4-nanometer process, which was installed on the X Elite platform. According to the plan, PC products equipped with Snapdragon X Elite are expected to be launched in mid-2024, and Oryon CPUs will also expand from PCs to fields such as mobile phones, cars, and wearables.
According to data disclosed by Qualcomm, the single core performance of Oryon CPU surpasses that of Apple M2 Max by nearly 14%, reducing power consumption by 30% under the same performance; Compared to Intel's flagship chip, the Core i9-13980HX, it can reduce power consumption by 70% under the same performance. The direct comparison also reflects Qualcomm's desire to compete on PC chips.
In fact, Qualcomm has already made attempts on PC chips and has released Snapdragon 7c, 8c, and 8cx based on the Arm architecture, hoping to break through with a combination of Windows and Arm. However, it did not cause a stir in the market until the rise of the Apple M series, which drove the ARM architecture's share in the PC market.
According to Counterpoint's data, in 2022, Apple led the ARM laptop market with a 90% share. Based on this, Arm's laptop market share has also increased from less than 2% to over 12% by the end of 2022.
In the past decade, ARM has been surging in the dark. Apart from Apple and Qualcomm, MediaTek, NVIDIA, and AMD are all ambitious. MediaTek has launched Kompanio 520 and 528 processors for entry-level Chromebooks; Recently, there have been media reports that NVIDIA and AMD are both developing PC chips based on the Arm architecture and will collaborate with Microsoft to support the Windows operating system. The product is expected to be released as early as 2025.
Under years of research and development, the competitive landscape is changing. Counterpoint predicts that from 2024 to 2025, Qualcomm and MediaTek's solutions are expected to achieve over 50% year-on-year growth on ARM laptops. By the end of 2027, Arm may eventually occupy approximately 25% of the laptop market share. Faced with competition, Intel, the largest manufacturer in the PC market, will lose nearly 10% of its market share within five years, but Intel will still dominate the PC market, occupying over 60% of the market share.
With the continuous rise of Arm's power, it has also brought new power to Windows PCs, and in addition, generative AI has opened up new space for the terminal side. At a 2022 earnings call, Qualcomm President and CEO Ammon predicted that Snapdragon Windows PCs would experience a turning point in 2024. In 2023, the PC market is indeed entering a new stage of development.
In addition to ARM chip assistance, on the one hand, the industry is expected to experience a recovery. According to IDC data, global PC shipments in the third quarter of this year were 68.2 million units, a year-on-year decrease of 7.6%, but a month on month increase of 11%, achieving two consecutive quarters of month on month growth.
On the other hand, AI PC has become a new theme, and today's press conferences from Intel, Qualcomm, and Apple all mentioned AI performance and the integration of AI related engines into new chips.
For example, the Apple M3 series has an enhanced neural engine that can accelerate machine learning models; Qualcomm Snapdragon X Elite has a Hexagon NPU that supports a large model of 13 billion parameters. When using Meta's Llama 2 model on the device, the processing speed can reach 30 tokens per second; Intel is about to launch its Core Ultra processor, codenamed Meteor Lake, on December 14th. It is also equipped with a Neural Network Processor (NPU) and provides a high-efficient AI acceleration experience on PCs. In future AI PCs, generative AI can be used to directly "automatically" compose music locally.
It can be seen that artificial intelligence has become the focus of chip manufacturers' efforts on the PC side. With the arrival of the AI PC era, a new challenge for computing platforms is underway.