The Science Behind Why The M1 Chip is So Fast
To start off, it’s important to note that the Apple M1 is a system-on-chip, not just a CPU. It contains a CPU along with other components including the CPU, GPU, unified memory architecture (RAM), Neural Engine, Secure Enclave, SSD controller, image signal processor, encode/decode engines, Thunderbolt controller with USB 4 support, and more.
THE COMPONENTS & WHAT THEY DO
The M1 has 4 high-efficiency cores that take care of less demanding tasks and use hardly any power.
Then, 4 high-performance cores that look after the more demanding tasks, and use more power in the process.
There are 8 GPU cores to handle graphics tasks. Things like drawing the interface and rendering 3D graphics for games.
We’ve also got a number of specialized processors for performing specific tasks. While a “normal” general purpose CPU core needs to be able to handle a multitude of different tasks, these specialized processors in the M1 are built to handle just a few. As a result, they are very power efficient.
Each time the system can offload processing to these specialized processors, it means less work for the CPU to handle.
As for what these specialized processors actually are. We have a video encoder and decoder which are able to convert video formats in a fast and power-efficient way. This is why the M1 can convert video codecs like h.265 without breaking a sweat, whereas an X86 machine (Apple’s previous CPU) will have to work a lot harder.
The M1 also features an image signal processor that speeds up tasks commonly done by image processing apps, including things like noise reduction in video.
We’ve got 16-cores dedicated to machine learning or AI tasks. To give an example of what this is, the iPhone’s ability to produce portrait photography by blurring the background and not the subject's face is all done thanks to this type of machine learning.
The M1 also has something called Secure Enclave. This is a high-performance storage controller that handles things like biometric authentication, encryption, and decryption tasks, and security. Again, the CPU is not having to do any of that work.
Next to the unified system, we have the RAM - Apple’s unified memory architecture that allows all of these cores, along with the CPU and GPU, to access the same memory. So, this is not a CPU, it is an entire computer on a chip. That alone is a key advantage Apple has because it can design and control the totality of the system, and then make easy tools that allow developers to make use of all of these features in a consistent way.
PC chip manufacturers and system builders simply can’t do that.
This is not to say Apple’s hardware is only quick because of that software optimization. The CPU cores are genuinely faster than equivalent X86 cores. In fact, these cores beat out pretty much everything from Intel.
HOW APPLE ACHIEVES THIS POWER
It comes down to the architecture. Apple has based their design on the ARM which they have always been interested in. They were actually equal shareholders with Acorn in the early ’90s.
Apple has been working on these chips for almost a decade.
ARM CPUs are RISC (Reduced Instruction Set Computers), whereas the previous X86 CPUs are CISC (Complex Instruction Set Computers. What I am trying to get at is the fact that there’s a fundamental difference in how these architectures approach computer tasks.
CISC vs RISC
CISC CPU’s do something called Out-of-Order-Execution. This means that the CPU can, to some extent, choose the order in which it executes instructions. This enables the system to complete more instructions at the same time which makes efficient use of the processor and as a result → speeds things up.
RISC CPU’s can also use Out-of-Order-Execution and the M1 absolutely does. Apple has focused very heavily on this and it’s implemented considerably more capacity for processing the underlying micro-operations than equivalent Intel and AMD CPUs.
In real-world usage, RISC turns out to be efficient.
If all of that still sounds complicated to you, it’s because it is. The main fact to know is that the M1 cores can process twice as many instructions as X86 cores at the same frequency. The M1 CPU cores are just genuinely fast and power-efficient, in a way that X86 cores simply aren’t.
Unlike other computers, the M1 Macs have their RAM mounted on the same package as that system-on-chip. In a traditional PC or Mac, the memory is mounted on the mainboard and then there’s a chipset that links it to the CPU.
Bringing all of this memory onto the system-on-chip speeds things up and it also allows all of the cores within the M1 to use the same memory. This increases efficiency and removes the need to copy memory from one area to another.
The M1 also uses the SSD for additional memory space, swapping out areas of memory that are not needed for a time so that more memory space is available for active tasks. Through clever use of memory compression and due to the speed of the SSD, you probably won’t even notice that this is happening.
It’s true that if you work the 8GB M1 model really hard for an extended period of time, you can get it to slow down but people have never been able to do that with the 16GB model. That really just shows how efficient this system is.
Whether these techniques will compromise the lifespan of the SSD — that remains to be seen, but there’s no doubt modern SSDs are incredibly resilient.
Combine all of these factors together, and it’s easy to see why the M1 machines are so fast. Sure they have their limits, but most general computer users won’t find those limits.
Of course, there are some minuscule issues with this first effort from Apple but considering this is only the first generation of the M1, these things will get fixed sooner than later. Future iterations of this chip have the potential to pull Apple way ahead of the PC competition in a lot of computing disciplines.