Intel’s 7nm Process Six Months Behind Schedule
How will this delay affect Intel's product roadmap, especially compared to AMD and TSMC?
Recently, things seemed to have been going well for Intel. Following years' worth of delays, the company finally got its 10nm chips on the market. But now the company has announced that it has faced issues with its upcoming 7nm process that will delay the release of its next-generation chips.
In Intel’s Q2 2020 earnings release, it was announced that “the company’s 7nm-based CPU product timing is shifting approximately six months relative to prior expectations….” This means that the originally planned delivery date of 2021 will now be pushed into at least 2022. However, if this delay mirrors previous Intel delays, it could be 2023 when the 7nm process finally hits the market.
What We Know About the Delays
According to Intel CEO Bob Swan, this delay has been caused by a “defect mode” that was identified in the 7nm process. This, Swan says, causes yield degradation issues.
As a result, Intel has begun investigating so-called contingency plans, which include using third-party foundries. Intel will also use external third-party foundries for its upcoming 7nm Ponte Vecchio GPUs, the company’s first graphics chips that come as a chiplet-based design.
Swan indicated that the GPUs will arrive in late 2021 or early 2022 pending a delay beyond the original schedule for a 2021 launch of the U.S. federal government’s Aurora supercomputer.
Intel's forecast for transistor innovation as of May 2019. Image (modified) used courtesy of Intel
Intel claims that the issues with its current 7nm process mean that production is trending a year behind its internal roadmap. Somehow, though, Intel says that being a year behind this roadmap will only result in a six-month delay in getting products to market.
Intel's Roadmap Compared to AMD and TSMC
Delays like this are nothing new at Intel. The company’s 10nm node also suffered years of delays because of problems at the foundry. And although its 10nm chips are now on the market, desktop 10nm processors are not expected to hit shelves until at least Q3 2021. In contrast, competitors like AMD have been outputting their own products based on 7nm architecture for months already, and some of these, such as the AMD Ryzen 4000, outperform Intel’s offerings.
Rival foundry TSMC is also on a steady timeline for new process nodes. In the same timeframe as Intel’s new schedule for 7nm, TSMC plans to be on the 3nm node.
Intel's process technology and packaging. Image (modified) used courtesy of Intel
Intel claims that it had a built-in buffer in its roadmap to account for process node delays, something that is most likely borne from lessons learned by the company in the wake of 10nm delays. Swan said that the company has “root-caused the issue” and believes that there are no fundamental roadblocks.
He remarks, "We feel pretty good about where we are, though we’re not happy. I’m not pleased with our 7nm process performance.”
Does Intel's 10nm Performance Compensate for the 7nm Delay?
At Intel's Architecture Day, the company pushed back against the idea that single-digit process nodes are inherently more effective than, for instance, a 10nm process. One of the major announcements at the events was a 10nm-based "SuperFin" (super FinFet) transistor, which is designed to increase performance in future Intel processors.
In an interview with Reuters, Intel's chief architecture Raja Koduri explained, "It is 20%, the largest intra-node jump ever in our history."
Performance leap in Intel's 10-nm process. Image (modified) used courtesy of Intel
"It’s actually the same as what you would get with one full Moore’s Law node of performance," he notes.
Featured image (modified) used courtesy of Amir Cohen/Reuters and Intel