For those who came in late, rumours claim that Chipzilla has difficulty getting 18A yield rates to acceptable levels.

Broadcom reportedly tested Intel's 18A process and was disappointed by the results. The primary issue was the yield rates, which have raised concerns, as low yield rates impede high-volume production.

The company's 18A process, touted as a breakthrough for the division, has faced criticism for its purported yield rates of less than 10 per cent, with the media labelling this as "abysmal.”

For comparison, TSMC's N3E process yield is rumoured to average around 84 per cent, occasionally nearing 90 per cent, while Samsung's 3nm process was reported to be below 20 per cent, a figure deemed too low for mass production purposes. A yield of 10% would similarly be considered insufficient for high-volume production.

Analyst Patrick Moorhead reported that Intel's 18A process wasn't tested on a PDK 1.0 but rather an older design kit, which might explain the low yield rates.

Gelsinger entered the discussion, corroborating Moorhead's claims and expressing pride in the "18A team" for their remarkable progress. He hinted that the reality on the ground differs from the mainstream media's portrayal.

In a reply to analyst Dan Nystedt, Gelsinger argued that assessing a process based on yield rate is flawed, noting that "a large die will have a lower yield, while a smaller die will have a high yield percentage."

Despite his departure, Gelsinger remains an ardent supporter of Intel Foundry, aspiring to make it the "world's largest" by 2030.

HotHardware points out that, while Reuters previously reported Broadcom to be dissatisfied with wafers etched on 18A, it is unconfirmed whether the company walked away from a semiconductor deal with Intel, as Chosun Daily suggested. Chosun Daily does not cite its sources regarding the 10 per cent figure, so a degree of scepticism is warranted.

Industry analyst Ben Bajarin tweeted in response to Reuters' report, noting that Gelsinger's previous statements indicated a 60 per cent yield range, which would be more suitable for mass production.

Gelsinger explained that Intel's chips had a defect density below 0.4D0 on 18A, corresponding to a 60 per cent yield range. He suggested that Broadcom's ability to port its libraries, rather than the 18A process itself, might be the primary issue.