The M1, the first member of the Apple Silicon family focused on laptops and desktop computers, is taking the battle that has been brewing for a long time right into the enemy camp. It is poised to pull down the curtains on CPUs as we have known them.
After nearly five decades — Intel 4004 came to market in 1971 — the central processing unit, aka the CPU, now has competition from System on a Chip, aka SoC. While a traditional CPU has maintained a stranglehold on the world of laptops and desktops, the SoCs have ruled the mobile world. And this detente was expected to persist, for no one thought that SoC could handle Intel’s best punch.
Here is how they are different. (Frankly these two images do a better job than my words.)
CPU needs an ecosystem of other chips to become a computer. They need memory chips for data, audio chips, graphics processors, connectivity chips, and more. SoC, on the other hand, as the name suggests, has everything on a single chip and is more efficient.
This is why mobile phones embraced the SoC approach to computing. Today, most computers are becoming derivative versions of mobile phones — lightweight, low power, instant-on, and always connected.
An SoC isn’t very much bigger than a CPU. However, when you apply cutting edge manufacturing technologies such as 5 nanometers, you can pack a lot more punch in an SoC. And Apple has done precisely that with its M1 — 16 billion transistors that do everything a modern computer needs to do. In comparison, a CPU still needs more chips around it to make a computer work — and that creates constraints for the machines.
With an SoC, there is a lot more room for disk storage and batteries. Because there are much higher integration and less internal wiring, power requirements are much lower, as well. You do the math: The sum of lower power requirements, fewer parts, and more room for batteries equals a machine that can last a day without a charge.
The SoC approach also means it is cheaper to build a computer. Richard Kramer of Arete Research estimates that an M1 costs somewhere in the range of $50 to $55 apiece. In comparison, a good CPU can cost between $150 to $200 apiece. And that is before adding memory and other chips. This is a significant opportunity for Apple to mop-up the lightweight laptop market — considering that there isn’t an x86 competitor in sight for at least a year. “We think Apple can increase sales of Macs by $18bn from FY20’s $28.6bn to $47bn, by growing units from 20m to 30m,” Kramer wrote in a note to his clients.
The most significant shortcoming of the SoC approach to computing is relative inflexibility. You can’t replace any components. There is no way to swap out CPU, GPU, or boost RAM. However, thanks to its tight operating system and chip-level integration, Apple can build custom SoC chips in various permutations. It can optimize the performance across its entire system.
As far as Apple is concerned, SoCs are the future of computing. Sure, there will be a need for general-purpose CPUs, but the writing is on the wall. Intel and AMD are embracing this trend of integration, though they are still selling CPUs.
Computing is changing, and so are the engines that power it. The sheer volume of mobile devices sold every year gives companies like Apple and Qualcomm a chance to better understand computing’s future. As a result, they can build chips for future computers better. I don’t see any difference between a 13-inch laptop and a tablet. And neither do companies like Apple.
For those unfamiliar with it, Moore’s Law — postulated by Gordon Moore, a co-founder of Intel — argues that the number of transistors on a microchip doubles every two years, while the cost of computers is halved. This has been the cornerstone of Intel’s success. The company’s ability to make the best chips before everyone else allowed it to maintain a hefty market share with outsized margins.
It is why it could afford to miscalculate and whiff on the mobile chip opportunity. It sold more expensive laptops, desktops, and server/datacenter focused chips, and it enjoyed Rolex-like profit margins. However, the company hit some manufacturing stumbling blocks — its transition to 10 nanometers and 7-nanometer manufacturing didn’t go as well. This ill-fated misstep happened just when mobile phones (and tablets) became dominant computing platforms.
Intel is still focused on the CPU and has no choice but to keep pushing ahead and seek more high-end CPU design opportunities. This means more advanced and complicated transistor designs, which lead to additional manufacturing challenges. So, now Intel is trapped and has to use outsourcing to make its chips. It is quite a fall for a company that once was known for its fierce chip independence and brutal approach to competition.
Intel’s chip manufacturing competitors, Samsung and TSMC, decided to bet on mobile and played it safe with the SoCs. The boom in mobile has enriched these companies. TSMC, for instance, is now making chips for Apple at 5 nanometers. They have a cash-rich client interested in the best manufacturing capabilities and will stay ahead of Intel’s curve.
Apple saw Intel’s challenges coming from a mile and has smartly moved away from the Intel platform. It knew that it had to build its laptop and desktop chips. M1 is the right first move. It is time to SoC the CPU.