M1 — Apple’s New Silicon Microarchitecture On A Chip
Apple innovates in many ways, whether it is the latest design or the best new features. They have taken the path to full vertical integration by developing some of their own hardware, including the processors. Apple began this move with the A series SoC (System-on-a-Chip) that first appeared on the iPhone and later the iPad. Apple also developed SIP (System-in-a-Package) chips which they introduced in the Apple Watch. These chips were designed using the ARM processor architecture, which differs from the x86 processor architecture used on Apple’s desktop and laptop models. Products like the Macbook Pro and the iMac use Intel-based processors that use the x86 instruction set. The PC is also using x86 processors made by Intel on models from Dell, HP and Lenovo (to name a few).
The move to an ARM-based microarchitecture for Apple’s desktop and laptop models is perhaps a big deal. It introduces what is called “Apple Silicon” to the rest of Apple’s retail line. It is also a move away from Intel, the primary chip maker for Apple’s desktop and laptop line since the departure from the PowerPC architecture. Ironically, the PowerPC was similar to the ARM, since both used a RISC-based instruction set. The move from PowerPC to Intel was mainly due to the lack of power efficiency, which affected Apple’s efforts to produce longer lasting laptops and less power hungry systems. It seemed unlikely Apple was going to return to that architecture moving forward, but ARM introduced a much better design which falls in line with Apple’s own principle.
During Apple’s One More Thing event last November 10, 2020, Apple made the official announcement of the M1 processor i.e. Apple Silicon. The first generation of products running on Apple Silicon will appear in the 2020 Macbook Air, Mac Mini and 13.3" Macbook Pro. The M1 is the first ARM-based processor to appear on Apple’s desktop and laptop line. This is also going to allow apps developed for the iOS to run on the devices running macOS, since they now have the same underlying architecture. No more different versions of an app since it will now be able to run across Apple’s ecosystem from an iPhone to a Macbook Pro.
Background
In June 6, 2005 during WWDC, then Apple CEO Steve Jobs announced the transition from the PowerPC to Intel. In his keynote speech, he made it clear what the reason for the transition was:
“When we look at Intel, they’ve got great performance, yes, but they’ve got something else that’s very important to us. Just as important as performance, is power consumption.”
The measure used to determine this is called performance per watt. This is in simple terms, how much performance can you derive from a product in a watt of power consumed. The product is referring to the CPU or processor (e.g. M1, i9–9900K, 5900X, etc.). The more that a processor can perform, the better. Unfortunately, the PowerPC did not have that capability. It was consuming more power to perform tasks that the Intel processor can achieve more efficiently.
We can measure efficiency in terms of performance per watt. This is basically saying if you can perform more tasks (operations) (t) per clock cycle (f) of the CPU by consuming less power (W), the better the performance in relation to efficiency (E).
E = (t / f) / WFast forward in time, and Apple is transitioning once more. Apple is looking for more efficient processors that can prolong battery life. There are benchmarks used to measure the performance which can draw observations in the lab. In the real world it also translates to the results of those observations, but not directly in the same manner due to other factors. This is because benchmarks are based on tests while users are more random in how they use a computer. The point here is that benchmarks are the best way to determine how the product will perform.
When the iPhone was developed, they did not consider using Intel processors because those chips are not energy efficient. Intel’s chips are primarily used for general purpose computing which run on desktop and workstations that are powered from the grid. What mobile device makers take into consideration are components that will run with less power at a longer duration using a battery. Since Intel’s chips consume more energy, it will not be ideal for battery powered mobile devices like smartphones and tablets. The solution was ARM-based processors like that from Qualcomm that are more energy efficient. Apple then developed their own which became the A series processors (e.g. A14 Bionic). The M1 is the next generation of Apple Silicon.
Introducing The M1
The M1 is Apple’s first generation of Apple Silicon SoC developed for computers (e.g. laptops, desktop). It is built by TSMC using a 5 nm fabrication process which fits 16 billion transistors on the chip. The design integrates the typical features found in a system into a single die on the chip. That includes the CPU, graphics, memory and advanced features into one component. This consists of 8 CPU cores and 8 GPU cores with integrated caches, neural engine and DRAM.
The CPU uses a BIG.little configuration which has a similar design to the A series chips. It utilizes 4 high-performance cores with 4 high-efficiency cores which balances workloads. All cores don’t have to run at the same power, which is why it consumes less energy. The high-performance cores deal with short burst compute loads that require faster clock speeds, while the high-efficiency cores handle background tasks or processes that don’t require as much power to finish. At the moment there is no official information posted regarding the actual clock speed of the cores, so that is something to look forward to.
The M1 uses a unified memory architecture or UMA, which creates a single pool for high-bandwidth and low latency memory access across all components in the chip. This improves performance since it decreases the data paths between the components to memory. This provides faster response times and improves the performance in graphics and video applications (e.g. streaming video and games).
According to Apple, the M1 boosts performance per watt up to 3x. This is in comparison to the latest Intel processors (not specific), so that means it can perform more tasks at less power consumed. This is perhaps a good reason to use the M1 for laptops like the Macbook Air and Macbook Pro. It will initially appear on the Macbook Pro 13.3" version, not the higher-end. This should give time for power users like editors and graphics specialists to see how well the M1 performs on the 13.3" Macbook Pro model. Sure enough, if that creates more demand for a higher-end version, it should appear on the 16" Macbook Pro as well but it may not be called the M1.
The integrated GPU, which has 8 cores, is another feature introduced in the M1. Apple has also developed their GPU without resorting to using the design from partners like AMD. If they did, then it doesn’t quite follow the vertical integration path that they are trying to accomplish. The M1 can execute 25,000 threads, which leads to Apple’s claims of having the fastest integrated graphics (based on testing conducted by Apple in October 2020 using preproduction 13.3" MacBook Pro systems with Apple M1 chip and 16 GB of RAM).
The Neural Engine on the M1 is perhaps one of Apple’s MVP features. This is capable of performing up to 11 trillion operations per second. It is built with a 16 core neural engine used for AI-based computations designed for machine learning. Apple claims it boosts performance by 15x, allowing compute intensive applications like video editing to perform faster based on these AI calculations.
Energy Efficiency
One of the main benefits of shifting to M1 from an Intel processor is the amount of power devices will save. An energy efficient chip design not only conserves energy for longer lasting battery life, but improves the performance per watt. Being able to execute more operations per clock cycle at the fraction of power consumed is very much needed as performance demands among users increases.
In the Information Age, users demand their smartphones and laptops to last longer from a single charge. Users are doing activities that now require more battery life, like streaming video and working remotely on location. There is also the inconvenience many users face with recharging their devices. While portable battery chargers can provide some additional power, there are times one is not available. There are also times when it is not possible to charge when there are no places available to plug in.
Another benefit to having energy efficient devices is to the environment. It creates a smaller carbon footprint when taken at scale of millions of devices. The idea here is that having to charge a smartphone or laptop less cuts costs and energy demand in the long run, so this requires less fuel consumed per user (these are estimates only and not based on a complete study).
Streamlined Ecosystem
It is also going to benefit Apple’s ecosystem as a whole. By having a common instruction set compatible with the underlying architecture, apps that run on one device can run on a different device. This allows the same apps that run on an iPhone iOS to run on a Macbook Pro, Mac Mini and iMac that are running macOS versions. Previously it would require having to download a different version of the app since iPhones use an ARM-based RISC instruction set while the Macbook Pro uses an x86-based CISC instruction set. Now that the next generation of Macbooks will be built with an M1 or ARM-based processor, it can run the same apps installed on an iPhone.
This would really be convenient for developers. Gone are the days of having to prototype for different CPU architectures. Instead they can focus on just one platform, reducing development costs and saving time in delivering an app to market. A software company would have teams developing an app for different platforms. It could also get complicated over time since different versions of an app are platform specific. A single platform simplifies the development process, but there might still be some details that need to be specific to an app (e.g. display on a smartphone vs. display on a laptop screen).
A more streamlined ecosystem can even lead to a single operating system. If that were the case, then there will no longer be an iOS, macOS, WatchOS or tvOS. Perhaps it will just be an AppleOS. The naming convention is appropriate though to distinguish the type of device. The naming could remain the same, but the operating systems on each device will derive from the same source that supports a common instruction set.
Less Dependency
Apple’s move to M1 lessens their dependency on other chip makers, notably Intel. While Apple still relies on chip manufacturer TSMC to build the product, the fact is the chip was designed by Apple’s own research and development team. They don’t owe any other company royalties or need to pay for licensing if they can build their own chip. With the M1 being fully Apple designed, if they had used AMD’s GPU like the RDNA 2, then they would have to pay AMD to integrate that with their product.
Less dependence on others gives Apple more control of development. In the past Apple would have to wait on Intel every time they needed chips. This proved to be a horrible relationship since Intel had problems fulfilling Apple’s orders. Apple had to rely on Intel’s product cycle in order to get the chips they needed, and it was affecting Apple’s own product releases. That meant Apple had to wait until Intel had the chips ready, but by the time they released it the product’s features would be obsolete or behind the times. That creates missed opportunities in the market, which can affect revenues. Apple has no control over that, which is why it makes sense from a business perspective to explore other options.
Final Thoughts
The M1 is just the first generation of Apple Silicon for the Mac line. Powerful computing no longer means power hungry systems that consume more energy to perform the most amount of tasks. The ARM-based processors show that optimized instruction sets that can use parallelism can improve performance at the fraction of the amount of energy consumed. This is the design philosophy Apple is adopting, as their products aim to boost performance with more efficiency that give users more battery time and value in terms of performance per watt.
Apple can part ways with Intel, a break up that was bound to happen after Intel’s production delays. Apple probably had enough of having to deal with Intel’s issues. Now that Apple has the M1, they have control over their own production cycle. Newer products with the best features can hit the market much faster than before.
While the M1 is for personal computing, I now entertain the idea that Apple could possibly develop a processor for gaming. Now that they can develop their own chips, perhaps the market in the gaming industry could become a possible target. For now the focus would be to get Apple Silicon on the higher end Macbook Pro models and iMac Pro. Then we can see it on a premium product like the Mac Pro. Let us also not forget that early versions tend to have issues, like bugs. Eventually they are ironed out, and only then can we see Apple Silicon expand to other products.
