The Intel 4 process, previously known as 7nm, is the first step along the path of recovery for the company’s logic technology development (LTD) group company after the disastrous 10nm process. Simultaneously, it is a necessary and critical milestone for the new “IDM 2.0” strategy that promises to regain Intel’s traditional semiconductor manufacturing lead over Samsung and TSMC and usher in a successful foundry business for third parties.
The Intel 4 process is not primarily a foundry node offering and is chiefly intended for a few designs that are high-volume but limited in scope, such as the compute tile in Meteor Lake. The Intel 4 Meteor Lake compute tile will be co-packaged with a graphics tile manufactured by TSMC, an SoC tile, and an I/O tile to form the basis of a complete client product portfolio that spans notebooks, desktops, and a few other markets. Since Intel 4 is tailored for a few designs, it eschews the rich feature set of a foundry process such as Intel 16 (previously 22FFL). As a concrete example, the process technology and design teams did not need to build high-density circuit libraries, which are necessary for graphics and many ASICs, but not for CPUs. Instead the Intel 4 process emphasizes high-performance, rather than maximum density. Limiting the focus and complexity of Intel 4 enabled the LTD team to focus on executing high risk changes such as aggressive design co-optimization, adoption of new materials, and EUV lithography ultimately delivering 20% better performance and doubling the density of the logic libraries.
Despite not being a full-fledged foundry node, Intel 4 is actually crucial to the company’s foundry strategy. The Intel 3 process (previously known as 7nm+) will be the first new foundry offering, alongside the existing Intel 16 node. It is essentially an enhanced variant of Intel 4, boosting performance and density through new process modules and other features, and will be used in the forthcoming Granite Rapids server processor. For example, Intel has promised more extensive use of EUV and faster transistors. Critically, Intel 3 will also bring a fuller and richer set of features to support a wider range of designs and customers, as appropriate for a foundry process.
In essence, Intel’s foundry ambitions rely on a clever two-step strategy to reduce risk and complexity across a process family. The first step is to drive the tightly focused Intel 4 node into high-volume production and achieve mature yield. The second step is creating the follow-on Intel 3 process by adding features to support more designs, improve density, and significantly enhance performance. According to the company, the Intel 3 node will improve performance by 18% over Intel 4. This is equivalent to the performance gain for a full node transition; comparable to the Intel 7 to Intel 4 transition and slightly greater than the TSMC 7nm to 5nm transition. By separating the ramp to high volume (Intel 4) from the development of a broader feature set (Intel 3), Intel will reduce risk and improve execution. This approach is quite similar to the collaboration between TSMC and its lead customer, Apple. Hence even though Intel 4 is most relevant for Meteor Lake and client products designs, it is absolutely essential to the success of Intel Foundry Services.