Apple brings the world’s first 3nm computer chip

In the development of computer processors, there is a very interesting period in history. From 2009 to 2017, Intel released a total of 7 generations of processors. Over the course of 8 years, the average performance improvement for each generation of processor was only 3% to 5%. During this time period, which is longer than World War II, users only saw a final performance improvement of 30%.

And at yesterday’s Apple event, it felt like I witnessed that ‘toothpaste era’ once again.

Yesterday at 8 am, the Apple product launch event took place online right on schedule. Within half an hour, they talked about numerous products, but in my opinion, Apple only released one product this time — the M3 chip, the M3 chip, and once again, the M3 chip.

The M3 series of chips introduced this time adopts the 3nm process technology, making it the world’s first computer chip to utilize TSMC’s 3nm process. With such advanced technology, within the cross-sectional area of a single hair strand, it can accommodate 2 million transistors.

Not only does the M3 chip feature an extremely advanced manufacturing process, but it also incorporates a new GPU microarchitecture. One significant breakthrough is the pioneering dynamic cache feature introduced by Apple.

Let’s compare the GPU to a “big eater” who needs to constantly eat, while the cache is like a conveyor belt filled with food (data) for the GPU to consume. Assuming the GPU can eat a maximum of 10 burgers at once for a given computing task.

Previously, the conveyor belt had to be set up to feed the GPU with 10 burgers at a time, even though it often doesn’t finish eating all 10 burgers. This resulted in the wastage of cache space.

However, Apple’s dynamic cache technology is like a conveyor belt whose width can be adjusted on the fly. When the GPU core needs to process more data, the conveyor belt becomes wider, and when it doesn’t need to process as much data, the conveyor belt width narrows accordingly.

This saves a lot of cache space, which can then be used by other GPU cores (up to 40 GPU cores in M3), resulting in significantly improved image processing capabilities compared to the previous generation.

In addition, the M3 series GPU also incorporates hardware-level ray tracing and mesh shading rendering from A17 Pro. This means that the new M3 chip has made significant progress in gaming and rendering, so it seems that the era when Macs couldn’t play games is about to become history.

According to Apple’s own presentation, in terms of CPU computing speed, the regular M3 chip offers a 35% improvement over the M1/M2, while the M3 Pro shows a 20% improvement over the M1. The M3 Max demonstrates a significant increase of 80% compared to the M1 Max and a 50% improvement over the M2 Max.

In terms of GPU computing speed, the regular M3 chip offers a 65% improvement over the M1/M2, while the M3 Pro shows a 40% improvement over the M1/M2 Pro and a 10% improvement over the M1/M2. The M3 Max demonstrates a 50% increase compared to the M1/M2 Max and a 20% improvement over the M1 Max.

Overall, the improvements in the M3 series compared to the previous generation are decent, with a general increase of around 15%. However, the M3 Pro lineup has left me feeling reminiscent of the ‘toothpaste era.’ It has transitioned from the previous design of 8 big cores + 4 small cores to 6 big cores + 6 small cores. Furthermore, the memory bandwidth of the M3 Pro has also taken a hit, with a transfer speed that is only 75% of the previous generation.

What’s even more surprising is that in the comparison, the M3 Pro completely skipped comparing with the M2 in terms of CPU performance and only compared it to the previous generation M1. This indicates that Apple lacks confidence in the improvements of the M3 Pro.

So this time, the M3 Pro not only feels like squeezing toothpaste but also gives a sense of toothpaste being sucked back in. On the other hand, the M3 Max, with its larger improvements, appears to have more sincerity.

Therefore, if there are individuals engaged in tasks that require heavy computational power such as rendering or video editing, it is recommended to switch to the M3 Max chip. As for those who primarily use their devices for daily office work, holding onto the M1 and continuing to use it is still a viable option.

Is Apple and Qualcomm about to dethrone Intel?

In addition to the updated processors, this time the MacBook also introduced a space gray version. It’s worth noting that the last time Apple released an all-black laptop was over a decade ago when Apple and Intel were still close partners. MacBook models at that time were based on Intel’s x86 architecture processors.
However, Apple has now progressed to its third-generation ARM-based PC processors.

Now, the question arises: why are giants like Apple and Qualcomm entering the realm of ARM-based PC processors? And how does ARM architecture compete with the long-established x86 architecture?

Before delving into this question, it is important to note that chip architecture and instruction sets are complex topics, and when using analogies to explain them, there may be omissions and errors. Your input and discussion are welcome.

Instruction sets and architectures are like the foundation of a house — they depend on each other. The instruction set is the blueprint of the foundation, while the architecture is the reinforcement and cement used to realize the blueprint’s functionality. On top of the foundation, you have the operating system, software applications like WeChat and WPS.

Intel and AMD, for example, represent the usage of the x86 instruction set. This blueprint is complex, and constructing the foundation is challenging, requiring more resources (power). However, it enables the execution of many complex instructions at once.

For instance, if you instruct it to build a wall on the foundation, you only need to provide one instruction, and it can complete the task.

On the other hand, Apple and Qualcomm use the ARM instruction set. This blueprint is simpler, and constructing the foundation is lighter, resulting in lower power consumption. However, the drawback is that it cannot execute complex instructions directly. To build a wall on this foundation, you need to give it instructions to lay the bricks first and then issue another instruction to use cement to connect the bricks.

Apple recognized the power-efficient and lightweight characteristics of the ARM architecture. They consolidated various computing units, such as the GPU, CPU, and NPU, into a single chip, forming their exclusive weapon — the M-series. When there’s a need for video processing, the GPU comes into play, while the NPU handles tasks like image processing.

Although traditional chip manufacturers like Intel can adapt to GPUs from companies like NVIDIA, the data sharing speed between them is far inferior to the integration achieved by designing them onto one chip. Coupled with the power-saving features of the ARM architecture, this integrated SoC (System-on-Chip) design is highly suitable for laptops that prioritize portability and battery life.

Software ecosystem compatibility is also crucial. Apple, unsurprisingly, has full control over the entire chain of technology, from chip hardware to application ecosystem, allowing for swift adaptation. Additionally, Qualcomm, as we discussed earlier, has received significant support from Microsoft. It seems that 2024 will be a transformative year for laptops.

Competition between ARM and Intel benefits consumers

With the invasion of the ARM architecture, Intel is feeling the pressure as the market is finite. The more market share Apple gains, the fewer devices are using Intel chips.

In response, Intel has been actively making changes in recent years. For example, their recent introduction of big core + small core designs aims to improve energy efficiency while maintaining performance.

However, for us consumers, competition among giants always brings better products. It’s always great to get a better product at the same price.

Regarding the new MacBook, the Pro lineup has discontinued the 13.3-inch model with a touch bar. In the future, we will only see the 14-inch and 16-inch models of the MacBook Pro. The new 14-inch Pro lineup starts with the M3 standard chip.

Regarding the previously mentioned M3 Pro chip series, the base model has increased in price from 15,999 to 16,999 yuan. The one thousand yuan increase resulted in a 2GB memory upgrade, once again confirming the rumor that Apple’s memory is as valuable as gold. The starting price for the M3 Max has also increased by 2,500 yuan, reaching 26,999 yuan.

The previous iMac all-in-one product has been upgraded directly from the M1 to the M3 chip, with a price of 10,999 yuan, which feels much more reasonable than the M3 Pro. For those who want to upgrade their iMac, this may be a good option to consider. These are all the products that Apple announced at yesterday’s event.

Finally, we can look forward to next year’s Windows notebook series. Based on the news from Qualcomm’s press conference, in 2024, we will see Snapdragon PC processors based on the ARM architecture. If the PPT is true, then it is highly likely that we will see a repeat of the situation where Apple’s M1 chip beat out Intel and AMD.