If you've ever looked at the spec sheet or an advertisement for a CPU, GPU, or even a fully built device like a laptop or desktop, you've probably seen hype around how it uses a 7nm or 5nm, or even 4nm process, node, or process node. But like many tech specs, the process node is much more complicated than a simple number, rarely explained by marketing, and not something you actually need to care about too much. Here's everything you need to know about process nodes, what they actually mean for computer chips.
Process nodes: a big reason why processors get faster every year without fail
Process nodes have everything to do with chip manufacturing, also called fabrication or "fabbing", which takes place in facilities known as fabs or foundries. Although virtually all chips are fabbed using silicon, there are different manufacturing processes foundries can employ, and this is where we get the term process. Processors are made up of many transistors, and the more transistors, the better, but since chips can only be so big, packing more transistors into a chip by reducing the space between transistors to increase density is a big deal. The invention of newer and better processes or nodes is the primary way of achieving greater density.
Different processes or nodes are differentiated by a length that has historically been measured in micrometers and nanometers, and the lower the number, the better the process (think golf rules). This number used to refer to the physical dimensions of a transistor, which manufacturers want to shrink when creating a new process, but after the 28nm node this figure became arbitrary. TSMC's 5nm node isn't actually 5nm, TSMC just wants you to know it's better than 7nm and not as good as 3nm. For the same reason, that figure can't be used to compare modern processes; TSMC's 5nm is totally different from Samsung's 5nm, and even in the case of TSMC's N4 process, it's considered part of TSMC's 5nm family. Confusing, I know.
New processes don't just increase density, however, they also tend to increase clock speed and efficiency. For example, TSCM's 5nm node (used in Ryzen 7000 and RX 7000 processors) compared to its older 7nm process can provide either 15% higher clock speed at the same power or 30% lower power at the same frequency, or a combination of the two on a sliding scale. Frequency and efficiency gains used to be much more dramatic up until the mid-2000s though, as shrinking transistors directly reduced power consumption in older processes, a trend called Dennard scaling.
The death of Moore's Law and what process nodes have to do with it
The key motivation for companies to use newer processes is to keep pace with something called Moore's Law, an observation made by legendary semiconductor figure Gordon Moore in 1965. The original law stated that the rate of growth for transistors in the fastest CPU is doubling every two years; if the fastest processor in one year has 500 million transistors, in two years there should be one that has a billion transistors. For over 40 years, the industry was able to keep up this pace by inventing new processes, each with higher density than the last.
However, the industry started hitting snags in the 2000s. First, Dennard scaling collapsed around the 65nm to 45nm mark in the mid-2000s, but after the 32nm process came out in the late 2000s and early 2010s, all hell broke loose. For most foundries, this was the last major node they would deliver for years. TSCM's 20nm from 2014 was simply bad and only its 16nm process in 2015 was a worthwhile upgrade from 28nm in 2011, Samsung didn't get to 14nm until 2015, and GlobalFoundries (spun off from AMD's fabs in the 2000s) had to lease Samsung's 14nm rather than make its own.
One notable exception to this turmoil was Intel, which successfully got its 22nm process out the door in 2011. However, Intel's release schedule and process quality started to slip after the 22nm mark. Its 14nm process was supposed to come out in 2013 but was released in 2014 with low clock speeds and high levels of defects. Intel's ludicrous goals with its 10nm node ultimately doomed it to development hell, missing its 2015 launch window. The first 10nm chip arrived in 2018, and it's one of Intel's worst CPUs ever. Intel's 10nm, renamed to Intel 7 for marketing purposes, wasn't completely ready until 2021.
The latest disaster concerns TSMC's 3nm node, which provides a significant improvement to density in logic transistors (which are what make up cores in CPUs and GPUs, among other things), but literally no improvement whatsoever to density in cache, also known as SRAM. Not being able to shrink cache is a total disaster, and it's possible foundries might run into similar problems on future nodes. Even if TSMC is the only fab that is struggling to shrink cache, it's also the biggest chip producer on the planet.
When you read about the death of Moore's Law, this is what it means, because if companies can't increase density year after year, transistor count can't go up. If the transistor count can't rise, then that means Moore's Law is dead. Today, companies are focused on keeping up with the performance implications of Moore's Law, rather than the technical ones. If performance doubles every two years, then everything's fine. AMD and Intel are using chiplets to increase both transistor count and performance while reducing costs, and Nvidia is relying solely on AI to pick up the slack.
Ultimately, process nodes are just one factor in whether a chip is good
Considering that a new process can make a chip smaller, give it a clock speed boost, and make it more efficient, all without making any major changes to design or architecture, it's obvious why processes are so important. However, other factors like packaging (such as chiplets or tiles or stacking chips) and AI are becoming increasingly viable ways to give value to a processor by boosting performance or adding features, not to mention simple optimization in software. The death of Moore's Law is unideal, but it's not the end of the semiconductor industry.
Additionally, because nodes are named for marketing reasons, there's no real reason to estimate a chip's competence based solely on its process; for example, Intel's 10nm is actually about as good as TSMC's 7nm despite 7 being less than 10. However, it's also true that a process isn't the only feature that matters in a processor. Plenty of CPUs, GPUs, and other processors have been bad despite being on good nodes, such as AMD's Radeon VII, which was a full process node ahead of Nvidia's RTX 2080 Ti and yet was so slow as to be one of the worst GPUs ever.
On its own, the process node of a chip doesn't mean anything. It would be like buying a CPU solely based on how many cores it has, or a console because it has blast processing. What really matters in a processor is its actual performance, which comes down to other hardware specifications and how well-optimized applications are for that hardware. If you're just wanting to know what the best CPU or GPU or laptop is, the process node won't tell you that. It just tells you who made the chip.