8 mins read

Intel reveals its next Intel 7, Intel 4, Intel 3, 20A and 18A nodes

Introduction

Pat Gelsinger, Intel’s brand new CEO, has just unveiled the company’s roadmap for its next 5 generations of manufacturing processes in what has undoubtedly been the most important announcement made by the company in recent months.

As we have revealed to you in this other news item, Intel has changed the way it names its manufacturing processes. In this way, Intel 7 is actually the new name they give to an improvement on the current 10nm node, which they are going to launch with the Alder-Lake CPUs. At the same time Intel 4 is actually the new name of the node that until now was known as 7nm and which they have been developing for years. Later, an improvement of that Intel 4 node will be called Intel 3. Later, processes 20A and 18A will come.

Let’s look in a little more detail at the expected improvements for each of the processes.

Intel 7 (previously known as 10nm ESF)

It is an improvement to the current 10nm SuperFin manufacturing process that allows them achieve 10% to 15% more performance per watt. They achieve this through transistor optimization, such as increased voltage, more low-resistance materials, new high-density patterning techniques, simplified structures, and better routing with taller metal stacks.

The first product manufactured in the Intel 7 process will be the long-awaited ones Alder Lake processors in late 2021 and the Sapphire Rapids for server in 2022.

Intel 4 (previously known as 7nm)

As we already knew (when we still knew it as 7nm) this is a whole new manufacturing process for which new factories and manufacturing technologies are being launched. It is Intel’s first FinFET node to fully use extreme ultraviolet (EUV) lithography technology..

The use of EUV means that they have to use an extremely complex optical system made up of lenses and mirrors that focus light beams of 13.5nm wavelength to print every tiny detail of the transistors on the silicon wafers. It is an important advance if we compare it with the optics of the lithographic systems used until now that emit ultraviolet light with a wavelength of 193nm.

All this in the end means that with this node they will be able to manufacture CPUs with a 20% performance per watt improvement regarding the Intel 7 process.

The Intel 4 manufacturing process will debut with the Intel architecture Meteor Lake consumer CPUs, entering production in late 2022 and the arrival to the consumer market in 2023. Granite Rapids will be the first server architecture to use this process, reaching the market the same year.

Intel 3

It’s an optimization over the Intel 4 process (which is probably why Intel just subtracted one number when choosing the name) with which They promise an 18% improvement in performance per watt.

The improvement is achieved thanks to the implementation of a denser and higher-performance library, an increased intrinsic current, an improved metallic stack for the interconnect with lower track resistance and a increased use of extreme ultraviolet lithography (EUV).

Intel has not revealed which architectures will use this manufacturing process, but They hope to start manufacturing the first processors in the second half of 2023. If we wait for the required time between the start of manufacturing and arrival on the market, we can assume that it will be at the beginning of 2024.

Intel 20A

This is where Intel takes a secular leap. The name change certainly reflects the importance of what they will implement. It consists of Intel leaves aside FinFET-type transistors (which were a revolution in 2011 and the most used today by the entire industry) for a new type of transistor they call RibbonFET.

RibbonFET is nothing more than the name that Intel gives to the new type of Gate-All-Around transistors (translated: gate around everything) or GAAFET. Both Samsung and TSMC have also been developing their own version of GAAFET for some time now. and in all likelihood they will overtake Intel in 2022. These transistors combine several silicon nanowires that are stacked vertically, making the gate that surrounds them much more compact.

Another important improvement that will be introduced in this manufacturing process is PowerVia technology. It is a power delivery system that It is done on the back of the chip. This avoids the need to route power from the front of the wafer. By getting power routing out of the way, they can allow for much more optimal routing of signals between transistors, while reducing skew and noise.

The Intel 20A process will be released in 2024.

Intel 18A

At the moment not much has been revealed about this manufacturing process. It is only known that they will use the latest EUV machines from ASML (Dutch company that provides lithographic machinery to all major foundries in the world), known as High-NAwhich are capable of obtaining even more precise photolithography.

Intel claims that they are the most important partner for ASML in the use of High-NA machines, and They will be the ones to receive the first production unit of a machine of this type. Thanks to these machines they will be able to develop and create much more advanced second-generation RibbonFET transistors.

2025 is the date Intel expects all of this to arrive, but so much can happen until then that they don’t include it in their new public roadmap at the moment. In any case, it is important that they mention it since It is in all likelihood the moment when Intel can once again lead the manufacturing processes of semiconductors, just as it did for decades.

Goodbye nanometers, hello Angstroms

Intel believes that nanometers are no longer an appropriate unit of measurement to call its manufacturing processes starting in 2024. The introduction that year of the new RibbonFET transistor and PowerVia technology should mean such an improvement in performance per watt that starting in 2024 that moment the number will refer to Angstroms instead of nanometers. Stick with the name because the word Ángstorm will resonate a lot in the coming years.

How much is 1 Angstrom in nanometers?

  • 1 Angstrom = 0.1nm
  • 20 Angstrom = 2nm

Therefore, Intel’s future 20A (20 Angstroms) node is equivalent to what a 2nm node could be. This will be followed by node 18A and from here we will see how the number decreases as they progress in the development of their manufacturing process.

Be that as it may, Nanometers or Angstroms, whether they end up leading again or not, the new CEO has suited Intel well. It seems that things are beginning to change for the better, that the company has a clearer direction, more focused and thirsting for leadership. As always, we will have to wait to see if the promises materialize.

Intel has ready the 20A and 18A manufacturing processes that will begin mass production by 2024

Intel shows the first wafer manufactured with the 20A process that will give life to Arrow Lake

Intel clarifies the new names of the next processors, the Intel Raptor Lake-S Refresh and HX will be named Intel Core ix

AsRock reveals its new Taichi Lite range motherboards for Intel and AMD

Intel Reveals the Specifications of its Arc A60 and Arc A60M Professional GPUs

Enermax reveals its Liqtech TR4 II compatible with Threadripper, Epyc and Intel Xeon processors

Intel discontinues its Intel Max 1350 Datacenter GPU after its January launch

Intel reveals Xeon 2024 CPUs: Sierra Forest and Granite Rapids!

Intel Reveals Aurora Supercomputer Specifications: More Than 21,000 CPUs and 63,000 GPUs

GIGABYTE registers several Z790 and B760 boards called X and a new YT PIONEER series for the next Intel