What is the M1 chip on the latest Macs?

 What is the M1 chip on the latest Macs?

Mac with Apple’s own processor. What kind of finish will the processor really have?

I was also paying attention to the Mac itself, but if I was looking for the “meaning of making it unique”, that attention would inevitably be focused on the performance of the processor. And the actually announced product further complements the Apple Silicon = Apple M1 chip for the Mac.

Apple has announced that it will double the performance of both CPU and GPU when the power consumption is set to 10 watts, with the expression “compared to the latest laptops”. Regarding CPU, it is said that “the same performance as the maximum performance of the latest laptop” can be demonstrated with 25% power for CPU and 33% power for GPU.

This latest laptop is a laptop with an Intel Core i7-1165G7, the latest model to use the 11th generation Intel Core and Xe graphics architecture. If their claims are accurate, they will have a significant advantage over other laptops in terms of performance per power consumption.

The appearance of the MacBook Air, MacBook Pro, and Mac mini announced this time is exactly the same as the previous model, isn’t it because it makes you feel the overwhelming difference from the Intel version? It’s a difference that makes me wonder. Even though they are all housed in the same housing, the MacBook Air has higher performance than the Intel version without a cooling fan. Even MacBook Pros and Mac minis with cooling fans have the scalability that the difference does not close. Had I changed the design, it wouldn’t have been so much different.

Now, let’s delve into the proof of each product in the actual machine review, but let’s focus on the “M1” here and think about what this SoC will bring to the Mac.

Why M1 is not A14X Bionic

As I did, I thought the first Mac would have an A14X Bionic, an enhanced version of the A14 Bionic with more CPU and GPU cores. However, a chip called “M1” has appeared. Perhaps the name A14X will be used for the iPad Pro, but there is actually a reason for the name M1 to be used.

That is, “M1 is not an extended version of the A14 Bionic.” The M1 is Apple’s latest circuit reconfigured and reorganized for the Mac. Of course, the latest circuits (components) developed by Apple are often the same as those developed with the A14 Bionic. High-performance core and high-efficiency core, 16-core Neural Engine, GPU core, ISP, etc. have the same design, but M1 has 4 same high-performance cores, 8 GPU cores, each in A14 Bionic On the other hand, it has doubled, and the memory bandwidth has been doubled as the GPU core doubled. This area is similar to the relationship between the A12 Bionic and the A12Z Bionic.

But Macs also require features that iPhones and iPads don’t need. One of them is virtualization, and the M1 adds functionality to run virtual computers (but not x86 processor virtualization, the instruction set is ARM). The GPU architecture is the same, but it supports more texture formats. This will be important in supporting games for PC and so on. Similarly, the A14 Bionic doesn’t have a Thunderbolt 3 / USB 4 interface.

In other words, we made the most of the design developed for the iPhone / iPad and used it as a component, but not all of it, but all the elements needed for the Mac in the same die. It is unclear whether the M1 will be used as it is when the iPad Pro is remodeled in the future, or whether the A14X Bionic will be made separately, but in any case, it is wrong that the M1 is a dedicated processor with elements unique to Mac. Absent.

Dynamically changing performance with thermal design. M1 has only “one” spec

All three Macs introduced this time are equipped with the M1, but with the exception of the entry model of the MacBook Air, which has a limited number of GPU cores, the specifications are single. The one installed in all models is M1, and the operating clock frequency is not listed there. This is because the maximum operating clock frequencies are all the same (because it is not open to the public, it cannot be known except by measuring it with an actual machine). To be precise, the operating clock frequency range is the same.

As with the latest 11th generation Intel Core, the clock that operates dynamically changes according to the cooling performance and operating environment (temperature, etc.) of the system equipped with M1 and the operating status of the application. In other words, the M1 itself is the same whether it is a fanless and thin MacBook Air or a mini desktop type Mac mini, and it can be considered that the instantaneous power is almost the same. However, if you have a game or application that is multi-threaded, has a full core, and runs the GPU to the full, it’s better to have a Mac mini with plenty of cooling capacity.

Given today’s processor situation, where performance is dominated by power, such a configuration is quite conceivable. However, from a different point of view, it can be said that Apple can use it in this way because it is a company that develops and sells personal computer systems.

CPU vendors such as Intel have so far selected the finished chips, classified their performance, and sold them at different prices for each rank. But for Apple, which sells only final products, it doesn’t make much sense to create a price lineup based on such selection, but rather it would be more natural for performance to change depending on the design.

High performance and restrictions due to shared memory SoC

Now, the reason why the M1 is power-saving and high-performance is that it introduces the integrated system chip (SoC) architecture that has been refined on the iPhone. In the SoC for iPhone, we have advanced the in-house development of the processing circuit to be incorporated, and have improved the perfection of the system architecture in which processing circuits for different purposes share the same memory and operate in cooperation. The shared memory architecture (UMA) is extremely efficient at running the CPU, GPU, and the machine learning model inference algorithm-dedicated processor (Neural Engine) in parallel.

Especially when using GPU for general purposes, it is inefficient to load data into video memory, process it, and draw it back to memory in order to reflect the result processed by GPU. It is better that the same data can be accessed by the processor that each is good at, such as CPU at one time, GPU at another time, Neural Engine at another time, and ISP at another time. It will be clear which is more efficient if multiple experts process the information around the same table or if the processing is carried around while exchanging emails.

Good efficiency also means that it is advantageous in terms of power consumption. The goodness of the design itself and the detailed power management function introduced in the A13 generation may have been successful, but UMA should also contribute to the reduction of power consumption. UMA’s advantage may be that when the processing capacity is improved by increasing the number of cores, the performance can be increased according to the number of cores simply by increasing the interconnection bandwidth inside the SoC.

However, there are restrictions because it is UMA. That is the installed memory capacity. The memory bandwidth shared by each processing circuit must be wide enough for UMA to perform well. It is presumed that the reason why M1 encloses DRAM in one package in the form of SIP (system in a package) is to maximize the bandwidth to memory.

Reasons to switch to Apple Silicon from the low-end side

In other words, the main memory can only be installed in the same package, so this time, the economically reasonable line would have been up to 16GB. In the future, whether or not the Apple M processor will be used in higher-end products will require an increase in the capacity of main memory.

Similarly, integrating the I / O interface into the chip is an advantage, but it can also be a weakness. The M1 only supports two Thunderbolt 3 / USB 4 ports. It has an interface inside the M1, but it’s expected to have up to two support channels. The concept of consolidating most of the system in one package is advantageous in terms of performance, power consumption, security, etc., but on the other hand, the hurdles for creating a large system are high.

For this reason, Apple has completely migrated the MacBook Air to the M1, but the MacBook Pro is only a lower model (Thunderbolt 3×2 port model) of the 13-inch model. The Mac mini will also be sold with an Intel-equipped model that has four Thunderbolt 3 ports.

There are also differences in the maximum memory and the SSD capacity that can be installed. If it is only simple CPU / GPU performance, considering the meaning of installing Neural Engine, it seems that it is okay to move to Apple Silicon more boldly, but part of the lineup, more specifically, in order from the lowest spec It is probably because there are restrictions on system expandability in all aspects, not just CPU / GPU performance.

Apple has said it will take two years to renew its lineup, and there is no doubt that it will take some time before the Mac Pro is replaced.

Sara Miller