How is this in any way relevant to the question of how easy it is to switch a fab line to new microarchitecture?I will point out that nobody ever talks about chip yields for particular chip designs, only for fab processes, e.g.:
![]()
Yield rate comparison of SMIC, Rapidus, TSMC, Samsung, Intel's advanced process
Comparison of advanced process yield rates between TSMC, Samsung and Intel: Numbers of advanced process yield rates Samsung’s advanced process yield rates are very poor and are not at the same level as TSMC. Perhaps we should say that TSMC is so good that it has brought the yield rate of...www.granitefirm.com
"For example, TSMC's 5-nanometer yield has risen from 50% to 80% in just a month after mass production, the 7-nanometer trial production rate is over 70%, and the 4-nanometer yield is 70%.
But for the 3nm part, the media has repeatedly revealed that TSMC's 3nm yield still has a lot of room for improvement. TSMC originally planned to mass-produce semiconductors for Intel and Apple with 3nm technology starting in July, but it has been unable to achieve the desired yield level. This can also be used to explain why the A15 Bionic SoC processor of Apple's iPhone 13 in 2021 is still using TSMC 5nm!"
Nobody ever talks about yields of particular chip designs, because yields are a fab thing, not a design thing.
You said switching to a different microarchitecture is simple, because it requries nothing more than a mask change. ["...you just make a new mask, and you can swap masks whenever you want. The rest of the fabrication process is the same."]
I replied I don't think that's correct because you'd need different metal layers to go along with the different architecture. I added making these layers is a complex process, so I think switching would be far less trivial than how you portrayed it.
Last edited: