This is another year when every chip foundry reminds us that they can make 3nm or 4nm chips, and we’re supposed to be excited that this technology is coming to the electronics we buy and use. But they never really tell us why. What exactly does it mean, and are there any real benefits?
Yes. But you probably wouldn’t notice them until someone tells you to look for them because it’s not the big deal chip makers want us to believe it is compared to other ways chips are getting better.
What is a nanometer?
A nanometer (that’s what the nm stands for) is a unit of metric measurement; 0.000000004 meters, to be precise. It’s tiny, so tiny you can’t see it without a microscope, and anything that small is extremely fragile, so a stiff breeze will break it.
When it comes to processors, the term nm refers to the node size of the chip. A smaller number means smaller transistors are placed on the silicon bed. A 6nm chip uses 6nm transistors, while a 4nm chip uses 4nm transistors. But not really.
It’s also a marketing term for processors. Yes, I’m telling you that a 4nm chip doesn’t actually have 4nm parts on it; they are smaller parts than the previously labeled 6nm chips, though (which weren’t actually 6nm, either). Here’s a video showing that a 14nm chip from one vendor uses parts the same size as a 7nm chip from another vendor.
Basically, it’s one of those things that only engineers will ever care about, but we’re told to care about it anytime a chip maker refines its process to use smaller numbers. Should we?
Potential benefits of a smaller node
There are two real, tangible benefits of a smaller process node: part density and energy efficiency.
You can fit more 4nm parts on a chip than you could 6nm. It’s not a one-to-one calculation, so you’re not actually getting 33% more transistors, but it will work out to a sizable number. That’s because the advancement isn’t just a numbers game, it’s more like an evolution that can be combined with other advancements.
Transistors are the worker ants of a processor and do the calculations. It doesn’t matter if those calculations are for AI acceleration, as part of a GPU, or just to tell when you tap a letter on your keyboard. If it requires any sort of processor activity (almost everything you do on your phone does), a series of transistors are doing that work.
In simple terms, more transistors can do more work. The process node advancement in other areas means the gains in performance can be significant. At least on paper, and we’ll talk more about that in a minute.
Smaller transistors also use less power. Like part density, other improvements will compound these gains and matter as much or more than any actual power savings from a smaller part. It’s important to remember that a by-the-numbers look means to see any actual improvements, 4nm transistors need to cumulatively use 34% less power than 6nm parts because there will be 33% more of them on a chip.
Again, it doesn’t work like that in real life, so the “other advancements” bit makes more of a difference when it comes to new chips being more efficient than the previous models.
This is why “4nm” is more of a marketing term than an actual measurement. As with any industry, chip makers like to toot their own horn about the process node size they have developed, but these numbers are mostly used because we can understand that 4nm is better than 6nm without knowing all of the details.
What about actual benefits for me?
The benefits of a smaller process are there even though we won’t notice them.
Yes, this goes against almost everything you’ll read telling why 3nm or 4nm is so great and such a breakthrough, but it’s the truth. That doesn’t mean that you shouldn’t care about it, but you should understand that this improvement is just part of the overall advancement of the modern processor and isn’t something you’ll ever notice on its own.
This is because there is a limit to any benefit of shrinking down circuits, and we’re very close to it. If chip makers only focused on building smaller transistors and circuits, they would spend billions to offer products that are no better than previous generations.
Even with all the other chip advancements, we barely notice a difference in a yearly chip upgrade, and those advancements can be much more dramatic than just shrinking the process node. Chip makers have to tell you that the latest offering has the potential for 12% better performance and 30% better battery life because you’ll never notice it if they don’t tell you to look for it.
The latest chip from Apple or Qualcomm isn’t great because it has smaller transistors, but it’s great because of all of the other “stuff” engineers are doing to improve performance and power efficiency. One day, we’ll see another breakthrough, but unless you compare a 4nm chip to something like a 28nm chip, size doesn’t really matter much.