Chips and nanometers: What is behind 7 nm, 5 nm, 3 nm and 2 nm?

If you find out about processors, i.e. the so-called “chips” in computers, smartphones and other electronics, you quickly stumble across the abbreviation nm. This stands for nanometers and has been used in relation to computer hardware since the 1960s. But what exactly does it measure and is the unit of measurement representing 0,0000001 cm perhaps slowly but surely obsolete? You can find answers and food for thought to these questions in this article. First of all: I am not an expert in physics or electrical engineering, so mistakes can be noted in the form of constructive criticism.

What do 7 nm, 5 nm, 3 nm and 2 nm mean for computer chips? What do the numbers say about transistors on processors like CPU and GPU? Here you get a (simple) explanation and an idea for alternative names.

What do 7 nm, 5 nm, 3 nm and 2 nm mean for computer chips? What do the numbers say about transistors on processors like CPU and GPU? Here you get a (simple) explanation and an idea for alternative names.

The current discussion about 5 nm and 3 nm

Will the chips in the new models of the Apple MacBook Pro still come with 5-nanometer or already with 3-nanometer architecture? This question has been discussed in recent weeks and months, and it currently seems that the further developments of the M2 chip (M2 Pro, M2 Max, M2 Ultra, etc.) will come onto the market as 5 nm models. 

This is probably due to the development status of the chip manufacturer TSMC, like the Apple analyst Ming-Chi Kuo in you Tweet dated August 22, 2022. Roughly speaking, the following can be stated for the discussion: the fewer nanometers, the more powerful the chip. In addition to the technology itself, the nm number is also important for marketing.

What does the tiny unit of measure on chips even indicate?

But what's the point? What will actually be displayed at the measly distance of 0,0000005 cm or 0,0000003 cm? Well, to put it very simply again: it's all about the spacing of the transistors on a chip. The smaller the distance, the faster and more efficiently information can be exchanged in the form of electrons. So the smaller the nm number, the more transistors per area and less resource consumption per computational effort are possible. The result is compact, high-performance components.

In addition to the distance between the transistors, the size of the chip itself is also important. Because the transistor density only really makes sense if you also know the area. In this way it can be calculated how many transistors are actually installed. More on that below. The mere specification of the nm architecture serves above all to say: "We are one step further, the chips are more efficient and (theoretically) create more computing power with less power consumption in less space".

The next level: chips with 2 nm architecture from IBM, TSMC and Co.

While with regard to Apple notebooks the question of 5 nm or 3 nm is in the air, the next generation of chips and transistor densities are already being addressed elsewhere. After all, the considerations outlined above are far from over. IBM already showed a 2021 nm chip as a prototype in May 2. Or rather: a 2 nm equivalent, because the measurement relates directly to the two-dimensional distance between computing units, but some semiconductors are now being developed using the more compact 3D process, which is why the nm designation is no longer so precise and comparable.

There is one at Ars Technica Post from May 2021, which deals with the topic. There you will also find an overview of the number of transistors per square millimeter on individual chips. Here is an excerpt for a better understanding and comparability of the components:

  • Intel desktop CPUs with 14 nm: max. 45 million transistors per mm2
  • Intel laptop CPUs with 10 nm: max. 100 million transistors per mm2
  • Apple M1 SoC with 5nm: max. 171 million transistors per mm2
  • Apple silicon with 3 nm: estimated max. 292 million transistors per mm2
  • IBM chip prototype with 2 nm: max. 333 million transistors per mm2

Advantages of ever smaller designs and smaller transistor distances

Before I try to use some half-baked information to argue for a new name for computer chips, let's look at the detailed advantages of the technical development behind it. Because in addition to all the theory and the race for the smallest nm number (regardless of whether it is real in 2D or as a reference in 3D), it is also important to understand why all these considerations and efforts exist at all. So these are the most important advantages of the ever smaller or closer together transistors on chips:

  • Processors can have more "cores" in less space and thus achieve more power while saving power
  • Information arrives faster over shorter distances, can be processed faster and consumes fewer resources
  • Smaller designs for chips and end devices in general are possible, so that mobile devices (iPhone, MacBook, iPad, etc.) are becoming increasingly powerful digital tools
  • Due to the lower consumption of resources, i.e. the energy-saving use, the heat development of the hardware is lower - less cooling is required and the technology is more durable
  • On mobile devices such as laptops, smartphones, tablets, etc., this all means longer battery life

DIY Optimization: Make Apple MacBook Air even more efficient

Is the nm chip unit still relevant?

As already mentioned above, it is not only the size and spacing of transistors in CPUs, GPUs and Co. that are becoming smaller and smaller. The physical structure of the chips themselves is also constantly being adapted. For this reason, some designations are only used as a reference in order to establish comparability with older technology. For people who do not deal with the subject in depth, this can not be understandable at first glance, and in the worst case can even be misleading. Perhaps, after more than half a century, we should think about a new nomenclature.

Traditional Labeling vs. Exact Quantities

Finding and using a new, more precise designation for semiconductors, chips, processors or whatever you want to call the hardware is of course not my idea. However, it fits well into the considerations here and has also been in circulation since April 2020. in one Post from HPCwire, in relation to a Paper taken by IEEE, useful information can be found. The article states, among other things, that scientists from MIT, Stanford University, the University of California/Berkeley and from the chip manufacturer TSMC in Taiwan would prefer a density specification.

The terms DL, DM and DC were suggested for this. DL gives the density of processor transistors in n/mm2 on. DM indicates the bit density of the main memory (currently the DRAM separate from the SoC), also in n/mm2. DC is intended to indicate the density of the connections between main memory and processor, also here in n/mm2. 

The linked article says about the application of the three values: "According to the authors, today's cutting-edge technologies can be characterized by [38M, 383M, 12K] on the basis of design specifics.So the proposed approach is not as promotional as a nm number, because you have to understand the meaning behind three different values. But it is technically more accurate.

Summary on computer chips and nm sizes

In conclusion, the size and spacing of transistors on computer processing units such as CPUs and GPUs are traditionally specified in nanometers (nm). And even if the two-dimensional unit of measurement is no longer really meaningful in modern chips, designations such as 5 nm, 3 nm and 2 nm still serve as comparative values ​​to show the technical progress behind the components. Because chips are becoming faster and more efficient. It is questionable what designations should come for even more advanced chips. There are solutions for this, but they are less easy to understand. 

-
Did you like the article and did the instructions on the blog help you? Then I would be happy if you the blog via a Steady Membership or at Patreon would support.

Leave a Comment

Your e-mail address will not be published. Required fields are marked with * marked