5 Comments

As lithos on twitter pointed out, it seems the original source isn’t the Korean press but actually New York Times which the Korean site just parroted -

https://www.nytimes.com/2024/12/02/technology/intel-ceo-pat-gelsinger.html

“TSMC is producing 30 percent of its leading-edge chips, known as 2 nanometer chips, without any flaws, while Intel’s new process produces less than 10 percent of its 18a chips without flaws, the person said.”

I agree with the other comment that the anonymous source is likely referring to Broadcom as it talks about “customers” which I doubt Intel product will be referred to as.

I think we can afford NYT some more journalistic credence. While we have no idea what the parameters of the 10% on 18A vs 30% on N2 is, one would hope that they know enough to verify that like things are being compared.

Given that, what do we make of it? One take is that BSPDN is more complicated so one would expect Intel to be a bit behind. But how much? The other possibility is that this is the reason why the board decided to fire Pat since it’s not doing as well as N2 so far and they have lost patience and don’t want more bad news to come out without them seemingly doing anything about it.

Expand full comment

Nice car analogy

Expand full comment

18A will decide the future of Intel. I think Intel don't need a lot of customers, just enough to cover the cost of Foundary and a little bit profit. If Intel succeeds, it will be back in the game, it will be AMD's worst nightmare.

Expand full comment

Regarding parametric yield. Is that more of a function of foundry or design teams ?

Expand full comment

I don't think "Panther Lake CPU core chiplet allegedly has overall yield of 10%." is the rumor!.

It's that Broadcom's test chip of unknown size had an yield of 10%.

Expand full comment