Intel has made its 14 nanometers a workhorse for more than
five years, exacerbating its know-how in the field to what we could define the
swan song of this production process . The 11th generation Intel Rocket Lake
processors will be the last to be based on this technology. As competition
gallops to increase density, the American chipmaker has never hidden his
reluctance to take part in this nanometer race.
Precisely in this context, Intel has blown everyone away
during the last Accelerated eventin which he unveiled his future plans, also
dwelling extensively on this concept and quoting, absurdly, the vice president
of the TSMC Research division : " Today these numbers are just numbers.
They are like the model in cars. It is just a nomenclature. to identify the new
node technology, so let's not confuse the node name with the potential that the
technology can offer ".
The new roadmap says goodbye to nanometers , with a change
of nomenclature as simple as it is important in terms of clarity also and above
all with the public. The nodes of the next generations will take the name of
Intel 7, Intel 4 and Intel 3 , respectively at 10, 7 and 7 nanometers. This
choice has enormous conceptual scope and aims to underline once again how much
the level of expertise reached by Intel allows proprietary 10-nanometer
processes to compete with the 7-nanometer ones of their adversaries. For that,
here's Intel 7.
Intel 7
The next hybrid generation of Intel processors will finally
adopt what everyone, insiders and non-professionals, has come to know as
Enhanced SuperFin at 10 nanometers .
Just a few days ago we talk about Koomey's Law and how it
can best represent technological progress in the coming decades.
Improved performance per watt will most likely be the key to
the future of the industry, albeit with raw power.
Already with Intel 7 we are in a range between 10 and 15% of
intergenerational improvement in terms of performance per watt , a clear message
of which is the direction of the giant of Santa Clara.
Alder Lake will therefore marry Intel 7 technology in the
consumer market . In parallel, Sapphire Rapids will represent its professional
counterpart. A first glimpse of the direction taken in improving energy
distribution and management, without sacrificing performance for this.
Intel 4 and Intel 3
The 7-nanometer production process will see the light with
Meteor Lake and Granite Rapids and here the discussion begins to get really
interesting. After an intermediate generation, which in the consumer field will
be represented by Raptor Lake, we are finally entering the era of EUV
lithography
The gain in performance per watt now rises by up to 20%,
doubling the density of Intel 7, with Foveros packaging technology and a 36
micron bump pitch. We are really close: the tape-in for the Meteor Lake CPUs
has already been there and the launch should take place by 2023.
Intel will arrive in quick succession 3. What for the most
mischievous could be nothing more than an Intel optimization 4, will actually
still use 7 nanometers but the focus will be all on density and energy
management. Not surprisingly, we are in terms of an additional 18% performance
improvement per watt. In this case, production should start in the second half
of 2023 but, in a scenario as realistic as possible, its arrival on the market
would be pigeonholed between 2024 and 2025.
The Angstrom Era
We are in 2024, in the so-called " Angstrom Era of
semiconductors ". Intel 20A will be the standard bearer towards the future
. Evolution and revolution, as we prepare to break through the nanometer wall.
The Intel 20A process comes dangerously close to us,
although in reality we are still in the 2 nanometer field, but the novelties
are many and essential.
On the one hand we find the RibbonFET technology , which
will eventually replace FinFET with an all new architecture of transistors with
gated on each side.
This new structure will improve the switching speed of
transistors, as well as ensure higher performance while physically occupying
less space.
Not a small leap, since FinFET has been with us for a decade now.
The other great innovation of Intel 20A is the introduction
of PowerVia technology, which aims to completely subvert the concept of power delivery
. Through PowerVia the energy will be conveyed directly under the transistors
instead of at the top, allowing not only a gain in terms of physical space but
also of energy dispersion, while allowing to reach higher frequencies. The
adoption of the PowerVia system will be made possible thanks to the Nano-TSV
technology , 500 times smaller than the current Through-Silicon Via.
Obviously, a look beyond the veil could not be missing.
Here, then,a nice Intel 18A teaser , but we'll have to wait a little longer to
find out more.
The evolution of packaging
If so far we have talked about production processes, Intel
has also allowed us to take a perspective look at future methods of packaging
and 3D stacking.
We naturally start from EMIB, the legendary 2.5D embedded
bridge technology from Intel , protagonist of both Sapphire Rapids and Ponte
Vecchio solutions.
Foveros and its subsequent iterations will instead be
adopted starting from Meteor Lake. In particular, in its first implementation
it will provide a unique 3D stacking methodology with wafer-level packaging
capabilities. Foveros Omni will succeed the latter by allowing greater
flexibility in stacking and aims to minimize the limitations of TSV technology,
while seeking to ensure high-performance die-to-die interconnections in 3D
stacking. An " agnostic " technology that also allows the combination
of tiles with decentralized manufacturing. Finally,
Foveros Direct will " blur the boundary between the end
of the wafer and the start of the package", with bump pitches of less than
10 microns and copper-copper interconnections to lower resistance.
