In recent years, Intel has often failed to meet expectations in terms of process progress. Even Intel 7, which is now widely used in Alder Lake, still encounters problems with Sapphire Rapids used in servers, resulting in delays in related products. According to Intel, it needs to span five process nodes within four years, and to reach Intel 18A in 2025, the pressure is very high.
Intel 4 is expected to be mass-produced in 2022H2
At the “Intel Accelerates Process and Packaging Innovations” in July last year, the latest process roadmap was announced, in which the Intel 4 process node (previously 7nm SuperFin) will be used by a variety of products including Ponte Vecchio, Meteor Lake in the client and Granite Rapids in the data center. Intel will use EUV lithography technology at this process node, which can use ultra-short wavelength light, about 20% performance per watt improvement and chip area improvement, and can apply next-generation Foveros and EMIB packaging technologies.
According to DigiTimes, an Intel source said that the Intel 4 process node is expected to be mass-produced in the second half of 2022. In fact, at the IEEE VLSI workshop about a month ago, Intel is optimistic about the progress of the next process node, saying that the chip can run more than 20% faster at the same power consumption or about 40% less power consumption at the same clock. It is understood that Intel 4 can provide double the transistor density compared to the existing Intel 7 process node (previously 10nm Enhanced SuperFin).
In addition to benefiting from the Intel 4 process node, Intel will also adopt a modular design for the first time on Meteor Lake for the consumer market, so that modules of different process nodes can be stacked, and then interconnected using EMIB technology and through Foveros packaging technology. Meteor Lake will also allegedly feature modules fabricated on TSMC’s N3 process.