Intel will no longer designate its processes with nanometers. The company is introducing new names that should give a “clearer picture” compared to competitors. The 10nm+ and the upcoming 7nm node will be named Intel 7 and Intel 4.
Intel 7 is the new name of what Intel previously called 10nm Enhanced SuperFin. This is an improved version of Intel’s 10nm SuperFin process used in Tiger Lake processors. Intel 7 is now in volume production and, according to Intel, offers a performance gain per watt of 10 to 15 percent over 10nm SuperFin.
This year, the Alder Lake CPUs for desktops and laptops will appear, made with the Intel 7 process. Next year in the first quarter, Xeon server processors of the Sapphire Rapids generation will follow. Earlier, Intel announced that these processors use the 10nm Enhanced SuperFin process, but that now has a different name.
Intel points out that since 1997 the naming of processes with nanometers no longer refers to the actual dimensions of the gate length. In practice, Intel’s 10nm process is broadly similar to TSMC’s 7nm process in terms of density. With the new naming, Intel seems to be clearly referring to that.
Comparison of three reference pitches, the minimum distance between two parts, in processes from Intel, TSMC and Samsung.
Intel renames its upcoming 7nm process to Intel 4. Intel will use euv for the first time in production. This generation should provide an increase of about 20 percent in performance per watt compared to Intel 7. Production on Intel 4 should start in the second half of 2022, for products that appear in 2023. These are Meteor Lake processors for consumers and Granite Rapids for data centers.
Completely new is Intel 3, where Intel will use more euv layers and further FinFET optimizations, which should result in an 18 percent gain in performance per watt compared to Intel 4. In the second half of 2023, Intel wants to start producing on this node.
Switch to ångström naming
Intel is also looking ahead to the future in which nanometers are no longer sufficient to indicate dimensions, according to the company. With Intel 20A, the manufacturer says it is entering the ångström era. A ngström equals 0.1nm, in fact Intel 20A is a ‘2nm’ process.
With Intel 20A, Intel also introduces new chip technologies, such as RibbonFET and PowerVia. The former is the successor to FinFET that Intel has been using for its chips since 2011, and PowerVia is a method of powering chips through the back.
Production with Intel 20A should start in 2024. After 2025, Intel 18A will come, with improvements to RibbonFET and use of High-NA-euv. Intel expects to be the first chip manufacturer to receive an ASML machine to make chips with this technology, the successor to euv. Intel says it is working closely with the Dutch manufacturer of chip machines for this. Intel wants to be leading in the chip industry again by 2025 with its 18A node.
Produce for third parties
According to Intel, it is “more important than ever” to provide a “clear picture” of its nodes, as the company will produce chips for other manufacturers with its Foundry Services. Intel thus assumes the same role as TSMC, whereby the company produces chips that it did not design itself. Intel announced those plans in March and has now announced Qualcomm and Amazon Web Services as its first customers.
With the new name Intel wants to compete with the current and future processes of TSMC. That Taiwanese manufacturer now produces a lot on its 7nm and 5nm processes and wants to start test production on its 3nm process this year. Mass production should start in 2022.