7 nm process technology in chips: Measure nanometers? Parsing

Snapdragon 865, Apple A13 bionic, AMD’s new Ryzen … Everywhere we are screaming about the 7-nanometer process technology in smartphones and PCs! How does this differ from the familiar 10 and 14 nanometers? How does it affect the battery, performance, heating? And then Samsung and Google are announcing 5 nm processors, someone is already talking about 3 nm.

Where is Intel in general? Just barely crawled to 10 nm?

We decided to find out what these nanometers measure? And is it so important to measure them or is it just marketing? And is Intel really so godlessly outdated?

Before moving on to the processors in our smartphones and computers, a few basics of how the processor works?

Meet – this is a transistor! A key element of all processors. In fact, a transistor is a switch. The current flows through it – this is 1, the current does not flow – it is 0. This allows us to count in a binary system – the basis of all processors!

Transistors used to be vacuum tubes. Conditionally – on or off: one or zero.

Such bulbs needed a lot to make it work somehow. For example, the 1946 ENIAC computer that participated in the creation of the hydrogen bomb totaled 17.5 thousand vacuum tubes and weighed 27 tons, occupying 167 square meters. At the same time, he ate 150 kW of electricity.

And here is one of the key points that you should pay attention to. Once again, I repeat the energy consumption of these 17.5 thousand light bulbs was 150 kW.

But in the early 1960s there was a revolution – the invention and the beginning of the production of field effect transistors. It’s just that their initial semiconductor is silicon – hence the well-known silicone, ahem, that is, Silicon Valley!

And then it started! The sizes of transistors decreased so much that they began to consume significantly less electricity and take up less space. And the number of transistors in computer technology began to increase at a tremendous speed! And with it the power of computing systems!

The first Intel 4004 industrial processor, which was released in 1971, had 2,250 transistors.

And now, for example, in the A13 Bionic of these transistors, 8.5 billion is more than people on the planet! Bye then…

But how much have modern transistors generally decreased, how small are they? A simple comparison is easy to understand – for example, with human hair!

Almost 1.5 million modern transistors made using the 7-nanometer process technology can be placed on its slice!

That is, you can place 4 times more transistors on the thickness of a human hair than there was in an Intel 4004 processor!

Why should it be reduced? Everything is more or less obvious here!

Firstly, the smaller the transistor, the less it consumes energy. You already understood this by the example of tubes.
And secondly – they are more placed on the chip, which means that productivity is growing. Double benefit!

And then we move on to the concept of process technology or Technology Node – what is it?

If you simplify it as much as possible, then the value of the process technology has historically been the minimum channel length of the transistor – as can be seen in the picture – do not confuse it with the dimensions of the transistor as a whole.

That is, the smaller the size of the process technology – the better – this is what the companies are trying to convey to us, but is it that simple?

And here something else is important: transistors are different and they differ not only in size, but also in their structure.

Classic, planar or flat, transistors have ceased to be used relatively recently – in 2012. They gave way to three-dimensional transistors, where they pulled the channel into the third dimension, reducing its thickness and thereby reducing the transistor itself. This structure is called FinFET – they are used now.

This technology has greatly helped to reduce the size of transistors, and most importantly, greatly increased the number of transistors per unit area, which is one of the key indicators for performance!

But does the concept of technological process today mean the same thing as several years ago?

A very important trend was observed throughout the industry – each subsequent manufacturing process was 30% less than the previous one, which helped to double the number of transistors while maintaining the same power consumption – for example, 130 * 0.7 = 90 nm, 90 * 0.7 = 65 nm, then to 45 nm, 32 nm, and so on.

And this so far complies with Moore’s Law:

The number of transistors placed on an integrated circuit chip doubles every 24 months.

What is behind this game of numbers?

We have already found out that the manufacturing process is the gate size of the transistor, that is, the length of the channel that passes or does not pass current through itself, and this size is key!

But this turns out to be true only if we are talking about the old 32 nm – everything is accurate there, at least measure with a ruler! And this parameter was documented!

But this was until 2009, when the concept of the technical process and its designation were excluded from the so-called “International Plan for the Development of Semiconductor Technology”!

In plain language – the numbers indicated in that process today – it’s just a marketing label!

The manufacturers peddled and began to call everything 10, 7, and generally 5 nanometers, and someone is already talking about 3 nanometers! You can put all this in quotation marks as a simple designation of the generation of processors!

Here’s an example of the structure of the Apple A12 processor manufactured at the TSMC factory using a 7-nanometer process technology. Pay attention to the scale bar in the lower left corner.

If we compare the scale and calculate, it turns out that the channel width is 8 nanometers, despite the fact that the process is officially called 7-nanometer.

Now let’s compare the 10nm process with Intel and the 7nm process with TSMC.

By the way, be aware that today TSMC is a company that manufactures processors for AMD, and also makes Apple A13 and Snapdragon 865 – so consider that we compare all their chips at once.

Pay attention to the dimension. It’s immediately obvious that the same 10nm for Intel is almost the same as 7 nanometers for TSMC! So Intel is not that far behind AMD and other manufacturers – did they just lose the marketing battle? Here, too, everything is not so clear!

Suddenly, in some respects Intel even outperforms TSMC.
Look at 1 square millimeter 10nm Intel chip fits about 5 percent more transistors than 7nm on the same Apple, Qualcomm or AMD.

But at the same time, the increased density also has disadvantages – increased heating!

So it turns out that Intel crystals are more powerful, but due to the density they are more heated. Thus, we get the same notorious throttling.

And the processors produced by TSMC – Apple Qualcomm and AMD benefit precisely due to the more spacious arrangement of transistors of approximately the same size.

How they do it is more a matter of internal architecture, rather than the number that appears in the name of those processes.

Do not think that I forgot about the architecture of N7FF + – yes it is even denser than that of Intel, but if we talk about chips series AMD Zen 2, Apple A13, Snapdragon 865 – all are based on TSMC 7FF and it loses in Intel density.

The only processor that is already manufactured using the new N7FF + technology using extreme UV lithography is the Kirin 990 5G. Here, of course, the density of transistors increases dramatically – as much as 15 percent!

In theory, manufacturers simply follow a slightly different path, and if you look into the future, it becomes clear which way: here is a sign of how everything will be – next-generation chips.

We are interested in a line about the density of transistors per 1 square millimeter!

According to these data, Intel bypasses both Samsung and TSMC in transistor density by more than 30 percent – and this despite the fact that here we are already comparing 7 nm from one manufacturer and 5 from another.

Where does this increase come from? How is such an increase in density possible – the process will simply explode or work only with a sophisticated cooling system?

Not certainly in that way. The thing is that Intel plans to switch to transistors of a completely different structure – under the name HNS – Horizontal Nano Sheets – this will make a leap!

But Samsung has similar plans – they go a little the other way to the Gate-All-Around FET structure.

That’s how it looks in reality – not so pretty, but just think about how small they are!

As a result, we realized that behind the marketing names of 7 nm and 5 nm there is a battle of architectures, and in the future we will be able to find out whose path was right.

What can be said for sure – a huge leap awaits us among all the chips, both mobile and desktop, over the next few years.

On this note, I do not want to finish the topic of processors, because we studied a lot of information and documents, including sorted out the production process. For example, have you heard of such an Extreme UV Lithography process? If on your fingers, this is some kind of fantasy – a drop of tin turns into a plasma after a laser hit: that’s how modern processors are created. But the installations themselves can only be created by one company in the world, and all the giants depend on it.

Similar Posts

Leave a Reply