TSMC 2 nanometer trial run yields exceed estimations at over 60%

[Redaktion]

Devices such as next-gen iPhones, Macs and iPad Pros are believed to be powered by processors supplied by the Taiwan Semiconductor Manufacturing Company (TSMC) as it has mastered the production of the bleeding-edge 2 nanometer production process. Now, the firm is said to be even closer to its target of producing this kind of silicon on schedule than expected.

https://www.notebookcheck.net/TSMC-2-nanometer-trial-run-yields-exceed-estimations-at-over-60.929028.0.html

[julia_top]

I think after the news that Intel and its 18A have a 10% yield unacceptable for any manufacturer which is possibly the reason for Intel's CEO firing more than 15,000 people and that Intel's foundries are delegating to TSMC hence the shortage.
TSMC's next-gen products are also currently projected to materialize as the A20 Pro for next-gen iPhones; MediaTek's future Dimensity 9600 flagship SoC; NVIDIA's Rubin series and AMD's Zen 6 platform.
More and more people and more OEMs are abandoning Intel distrusting their products and distribution.

[MXM0]

Careful with these numbers -- TSMC is self reporting here, and their test chips might be tiny.  Failure rate is really more about flaws per unit area, than an abstraction of 60% on "test chips of unknown size".  TSMC also has a rep for pushing the bounds of truth a bit -- just like with their initial N3 node.

[MXM0]

I think after the news that Intel and its 18A have a 10% yield unacceptable

That report was in September, and it was from Broadcom, who were specifically testing 18A as a platform for AI chips, which are huge, and prone to much higher fail rates, because of it.  I suspect the failure per unit area here is much closer than it seems, given that the Intel numbers are several months old now, and the test chips were almost certainly AI (big) vs mobile (small), which would have yielded this kind of difference by area alone.

[idontknow]

These articles make no sense, they are basically juggling meaningless (and outdated) numbers around. Please Notebookcheck make some research and actual journalism instead of regurgitating the same news articles that the other sites are posting.

TSMC's yield is 60% at what chip size?? That very same 10% yield on Intel's node could result 99% yield rate with a chip small enough. Im really starting to think there is some kind of smear campain around intel by someone or a group of people trying to get their stock for cheap.

[Davidasdzxc]

Guys, use some common math, even if TSMC used the 60% on reasonable small chips, we could say that this is indication of flaws in area

[Hotz]

TSMC's yield is 60% at what chip size?? That very same 10% yield on Intel's node could result 99% yield rate with a chip small enough. Im really starting to think there is some kind of smear campain around intel by someone or a group of people trying to get their stock for cheap.

The problem is that Intel constantly had problems with smaller nodes in their own chip manufacturing. Everything smaller than 14nm had bigger issues, and didn't get enough time to mature. Why should it be different on a newer, even smaller node?

It will have problems. The only question is how much? If these 10% yields are wrong, Intel still gave us no numbers what's the their yield percentages. They only said "it depends on chip size", which neither confirms or nullifies the 10% claim.