So you’ve heard about Intel’s new 10nm processors, or perhaps about Apple’s 7nm A12X Bionic ones, and you’re aware of the GHz and core count, but you’ve no idea what those “nm” numbers signify… No worries, let us explain.
What does the “nm” mean?
A CPU that you and I use in our daily lives, whether in a mobile phone, or a smart speaker, or a laptop, is made up of transistors. Billions of tiny transistors, that are basically electrical gates that switch on and off to perform various calculations. The “nm” in these numbers we’re talking about stands for nanometers – a minuscule unit of length and this “nm” parameter denotes the size of these small transistors that the CPU is made up of, and the distance between these transistors. This parameter is also known as the “process”, the “process technology”, or the “CPU lithography”, and the process of manufacturing these CPUs is known as “Fabrication”.
For reference, Intel’s 10nm processors that debuted last year are built using 10nm transistors, and similarly, Apple’s A12X Bionic and AMD’s new CPUs are made using 7nm transistors.
Lithography history, present, and future
Back in the 2000s, desktop processors used to have 42nm processes, which sounds too large in comparison to today’s processes. Then came 28nm process nodes, and then came the 22nm processes. In the 2000s, transistor size shrunk in size by half almost every 2 years. This shrink had almost ceased since 2013 for desktop processors. Then in 2018, Intel announced their 10nm Sunny Cove chips, and AMD soon announced their smaller 7nm CPUs (based on TSMC’s 7nm process) soon after.
The same goes for smaller processes as well. Mobile CPUs have come down to 7nm (Apple A12X Bionic) from 14nm in the past decade. As of January 2020, mobile CPU biggie Qualcomm is already on the 7nm process for the majority of its new CPUs, and Apple is said to ship phones with 7nm CPUs for at least a couple of years. Honor also has shifted to the 7nm train with their Kirin 980 CPU. Samsung also announced its Exynos 9825 CPU in 2019, which was the first mobile CPU built on the 7nm process using EUV (extreme ultraviolet lithography). Similarly, MediaTek also announced a 7nm CPU in 2019.
Update: Apple has shipped the A14 Bionic SoC with 5nm process in the new iPhone 12 series on 13th October 2020. This makes A14 Bionic the first 5nm SoC ever on a production-ready mobile phone. The second one would be Kirin 9000 SoC that debuted in the Huawei Mate 40 Pro on 22nd October 2020.
7nm is currently the market trend, and the norm that PDAs will follow for at least a couple of years from now, but TSMC and Samsung are already said to be working on 6nm, 5nm, and even 4nm processes.
Why are smaller processes important?
The reason why everyone wants smaller transistors is that they’re better (duh). Firstly, smaller transistors will occupy lesser space physically, which means you can fit-in more transistors, which in turn means more powerful processing. Secondly, smaller transistors will consume less energy, which means lower power consumption. And thirdly, because of lower power consumption, there’s lower heat dissipation, meaning cooler processors.
Mobile chips will see the biggest improvements with these smaller processes. Not just in terms of performance, but also in terms of power efficiency. On paper, a 7nm chip can deliver 25% more performance than a 14nm for half the power consumption. We have already seen Apple’s A12X crush many desktop processors in benchmarks, despite being passively cooled.
It isn’t as easy as it sounds though – the process of shrinking these processes. Making smaller transistors requires very precise instruments and machines, which is why processors built on smaller processes will be costlier than older, larger ones.
As we see brands race towards 4nm, the prices shall see a normalization. Samsung is very confident in its EUV roadmap and aims to be the first to announce the 4nm process.
But, what after 4nm? 2nm? Or what after 1nm? Well, how much tech “shrinks” is for all of us to witness, and with time, we will. What I can expect for now is that this article helped you learn something new. Until some other post or video (and until processes shrink beyond 1nm), peace out.