News about the high defect rate of Intel 18A turned out to be fake: the ex-head of the company explained why

In the middle of last week, the Korean publication The Chosun Daily casually stated that the yield level of usable chips using the Intel 18A process supposedly does not exceed 10%, and this scares away potential customers from the technology. The more popular resource WCCFTech inflated this topic to a magnitude that could no longer leave the former head of Intel indifferent, and he had to refute the rumors.

Image Source: Intel

As a matter of fact, many industry experts, including the famous Patrick Moorehead, managed to speak out against such assessments of the situation last week. The latter’s comment, branding the rumors of South Korean origin “fake news,” drew the attention of former Intel CEO Patrick Gelsinger, who was suddenly relieved of his post. He thanked his namesake for his commitment to defending the truth and added that he was “very proud of the incredible work and progress that the developers of Intel 18A technology are demonstrating.”

Note that at all recent events with the participation of Intel representatives, they did not tire of repeating that the company’s own products manufactured using Intel 18A technology will be launched into mass production next year, and the company already has working samples of individual Panther Lake-type processors. If the product yields were so bad, Intel would not be able to prepare for the start of mass production in 2025.

There has been a growing debate among industry experts that product yield levels are often either confused with the defect density on a silicon wafer, or the two indicators are not linked. The latest trend, as it became clear from Patrick Gelsinger’s comments on the pages of the social network X, upsets him especially greatly. He explained that it is unacceptable to talk about the yield of suitable products in percentage terms: “A large crystal will give a lower level of yield of suitable products, a smaller one will give a high level.” It is incorrect, according to the former head of Intel, to talk about the yield of suitable products without reference to the crystal area of ​​the manufactured products. It sounds logical: the chance of “messing up” in a large area is much higher than in a small one.

Against this background, Taiwanese media report decent quality indicators for the first products produced by TSMC using 2nm technology. According to their data, the yield level of suitable products in this area exceeds 60%, and this predictably pleases TSMC representatives. Such messages prompted the former head of Intel to stand up for his native company in a situation of escalating rumors about problems with the development of the Intel 18A technical process.

It is not without reason that Taiwanese TSMC is starting to use new technological processes in the production of compact mobile chips, since with an equal density of defects they produce a lower percentage of defects. Intel, on the other hand, is focused on producing larger chips, and its defect rate may be higher, all other things being equal. However, the real quality indicators behind all these discussions remain undisclosed, and Intel, in a difficult time for the company, has no particular motivation to discuss them openly.