Another day, another “bad news for Intel” story. In a year of terrible financial results, disappointing desktop chips, internal turmoil and… you know the rest. However, if the latest rumors are to be believed, the situation in the foundry industry may be even worse than previously thought.
Following the resignation of CEO Pat Gelsinger earlier this week, a report from the South Korean outlet was released Chosun every day they say that Intel’s 18A process provides only 10% efficiency (via TrendForce). Gelsinger has previously stated that he would “put the entire company on 18A,” and that seems like a risky venture that could prove disastrous if this latest data turns out to be true.
However, Intel stated this back in September “18A is on, healthy and yielding well.”leading some to speculate that the reported low yield rates are faulty.
To give you some context right off the bat, efficiency is measured as a percentage of the total number of chips that can be produced from a silicon wafer.
According to rumors, the performance of TSMC’s N3E process is average about 84%sometimes approaching 90%, while Samsung’s 3nm process was like that reported to be below 20%which was considered too low for mass production purposes.
A yield of 10% would probably be considered very similar. However, – emphasizes HotHardware that Broadcom had previously been reported by Reuters dissatisfied with the waffles Etched on 18A, it has not been confirmed whether the company has walked away from its semiconductor deal with Intel, as reported by Chosun Daily in the same article.
This would raise doubts about the suspiciously low yield percentages reported here. Chosun Daily does not cite sources for the 10% figure, so a pinch of salt will most likely be needed.
In reaction to the Reuters report, industry analyst Ben Bajarin he tweeted that Gelsinger’s previous statements showed an efficiency range of 60%, which would be a much more appropriate value for mass production potential.
Gelsinger pointed out that Intel’s own chips had a defect density of less than 0.4D0 at 18A, which is within the 60% efficiency range. Broadcom’s ability to port its libraries is probably the main issue, not necessarily 18A. Which is ironic because Broadcom once boasted… https://t.co/Q2lFP82qYUSeptember 4, 2024
Still, Intel’s foundry business has been an ongoing problem for the company as it has abandoned the 20A process altogether in favor of focusing on an 18A-based future. Next-generation Nova Lake and Panther Lake processors are expected to be built mainly on 18Aa disastrous profitability rate would put Intel even further behind than it currently appears.
Time will tell, I guess. The current generation of Arrow Lake chips were built primarily by TSMC, and while they proved to be somewhat of a disappointment, Intel hoped that the 18A performance was great enough to mean that its future processors would be much more of an in-house creation. romance.
That will happen if the chip giant doesn’t cut itself into too many pieces by then, or if the U.S. government doesn’t stick its nose in to force the company to split up. Our Jeremy wrote a fantastically detailed article on what could happen after Gelsinger’s departure with Intel and its chip business, but the bottom line is that things are looking pretty uncertain for Team Blue in the near future.
What if these rumors are to be believed? 18A can contribute significantly to this bumpy ride. Chip production, right? It’s a tricky business, even when it comes to tailgate sorting.
