You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Thanks for your answer. I see now chiplet is a wrong word. Perhaps die is a better word? I do not see any dies on the wafer. That was what I was trying to say.
UMC February revenue also dropped by about 8.1% from that of January if one looks carefully at: https://semiwiki.com/forum/index.php?threads/umc-reports-sales-for-february-2025.22242/
1-18193515/(38000310-18193515)=8.1%
That rounded thing is actually more of a demonstration of how many chips are packed into an NV72 rack, ie, if all the chips in the rack are taken out and rearranged in a circle. Such a big circle cannot be fab by anyone, but is nonetheless already realized in the rack.
It is a demonstration of...
The "Ascend 910B" is a very large and unusual chip. It is hard to conceive TSMC got duped to fab it unwittingly.
It simply does not make sense for them to provide a cover for Huawei at the risk of getting their reputation damaged, or exposing themselves to the wrath of US sanctions.
The legal filing is some 40 pages long. It's a good read on the development of OpenAI. If you read it, you may become impressed with what Elon has contributed to the advancements in AI, in addition to his other achievements.
Link: YMTC 232-layer 3D NAND memory: an unexpected technological breakthrough – The chronicles by Yole SystemPlus
Are we witnessing some unexpected consequences of US sanctions?
The original text is "above 14nm", lacking the word indicating "inclusive", thereby entertaining the ambiguity of whether 14nm is included. Not sure if that is a big deal though.
