Intel has entered the angstrom era with 20A and 18A process nodes
Intel CEO Pat Gelsinger took the top job at the semiconductor design and manufacturing company in February 2021. A big part of his remit was to manage a turnaround, including the ambitious five nodes in four years (5N4Y) strategy announced in July 2021.
In addition to supporting Intel’s own product roadmap, 5N4Y was meant to bolster Intel Foundry Services, announced in March 2021, which would see the semiconductor giant evolve its integrated design/build model to open up manufacturing capacity to fabless competitors.
Gelsinger said in a statement when 5N4Y was announced that the company is “accelerating our innovation roadmap to ensure we are on a clear path to performance leadership by 2025. We are leveraging our unparalleled pipeline of innovation to deliver technology advances from the transistor up to the system level. Until the period table is exhausted, we will be relentless in our pursuit of Moore’s Law and our path to innovate with the magic of silicon.”
While Intel is currently facing numerous macro headwinds, in part related to its position in artificial intelligence (AI) relative to AMD, NVIDIA and others, development of U.S.-based semiconductor manufacturing capacity is viewed by many, particularly many in the U.S. government, as a national security issue, important in the context of rumors that Intel would spin-off foundry. In fact, the company recently reorganized foundry as an independent subsidiary within Intel.
Subscribe now to get the daily newsletter from RCR Wireless News
The logic, as Gelsinger explained in a Sept. 16 message to employees, is the spin-in “provides our external foundry customers and suppliers with clearer separation and independence from the rest of Intel. Importantly, it also gives us future flexibility to evaluate independent sources of funding and optimize the capital structure of each business to maximize growth and shareholder value creation.”
Back to the process and packaging strategy; these are the process nodes covered by 5N4Y:
- Intel 7 is a 10-nanometer process—prior to being called Intel 7 it was called Intel 10-nanometer Enhanced SuperFin (10ESF)—used for the Alder Lake client CPUs and Sapphire Rapids for data center.
- Intel 4 is a 7-nanometer process that introduced Extreme Ultraviolet (EUV) lithography, which replaced deep ultraviolet (DUV) lithography, used for Meteor Lake on the client side and Granite Rapids on the data center side; it also introduced the chiplet approach to packaging wherein modular integrated circuits are inter-connected as opposed to a monolithic chip combining all functions.
- Intel 3 is a 3-nanometer process that builds on Intel 4 by improving performance per watt by 18% as compared to Intel 4. This one is focused on data center use cases and productized as Xeon 6, which includes performance and efficiency cores for client and data center products.
- Intel 20A marks a switch from nanometer-based measurements to angstrom-based measurements; there are 10 angstroms in one nanometer. This process introduced backside power delivery and Ribbon Field-Effect Transistor (FET) architecture, also called Gate All Around (GAA) FET. But according to a blog from the company, learnings from 20A “have directly informed the first commercial implementation of both technologies in Intel 18A.” The Arrow Lake CPU will be fabricated by a third-party and packaged by Intel.
- Intel 18A, as mentioned in the above bullet, includes the backside power deliver and RibbonFET transistor architecture, and is “on track” for a 2025 launch, according to Intel. “Focusing resources on Intel 18A also helps us optimize our engineering investments,” company VP of Technology Development Ben Sell wrote in a Sept. 4 blog. “When we set out to build Intel 20A, we anticipated lessons learned on Intel 20A yield quality would be part of the bridge to Intel 18A. But with current Intel 18A defect density…the economics are right for us to make the transition now.”
So if the 20A process node was built by Intel but the chips to be used in Arrow Lake products were built by a third-party, and acknowledging the relationship between 20A and 18A, did Intel deliver five nodes in four years? Seemingly, yes. But we asked them for their take.
“We’ve always viewed Intel 20A as a bridge to Intel 18A, and the advancements we achieved in Intel 20A are feeding directly into Intel 18A,” a company spokesperson told RCR Wireless News. “We are encouraged by what we’re seeing from Intel 18A in the fab—it’s powered on and booting on operating systems, healthy, and yielding well—and we remain on track for Panther Lake production in 2025. The health of Intel 18A is why we’re concentrating our engineering resources to complete 5N4Y and continue innovating on future process nodes.”