Ouch, ouch, ouch, the news from Intel and its engraving processes is still not good. While the company is betting heavily on its 18A, it would appear that the latter is still not up to scratch. In fact, the latest reports on the subject point to catastrophic performance, making it clearly not ready for mass production… And to think that the blues have cancelled their 20A to concentrate on this one.
18A: one chip in ten viable!
From what we’ve learnt, Intel is having major performance problems with this new node etching. And yet the company has bet big on it, notably by cancelling its 20A, but obviously that’s not enough. We have learned that the company’s performance with this process is a catastrophic 10%. To put it crudely, on a silicon wafer, only one chip in 10 is viable; the rest either fail qualification tests or end up in the bin.
With yields like these, it’s easy to see that the economic consequences are considerable. Broadcom has expressed its fears and disappointment on this subject. It should be noted that the company opted for this process (before cancelling its orders, according to Chosun) for the manufacture of its future high-bandwidth network processors.
In short, this latest failure could also be one of the reasons why Pat Gelsinger was ejected from the Intel board. It should be remembered that he wanted to turn the company into a major semiconductor manufacturer like TSMC. But with the disastrous performance of 18A and significant financial losses, he had no choice but to leave.
It now remains to be seen what the consequences will be for the company. Clearwater Forest’s Xeon processors for servers were due to use this etching by 2025. If it is not ready, adjustments will have to be made so that this generation is produced by TSMC.
