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!
Now that Samsung has been outed for renaming their 3nm to 2nm, those whales are probably wondering how they’ll fit those huge knee pads under their slacks, when they return to TSMC to grovel.
Now that Samsung has been outed for renaming their 3nm to 2nm, those whales are probably wondering how they’ll fit those huge knee pads under their slacks, when they return to TSMC to grovel.
The current 3nm is SF3E, then SF3 and SF3P and there were rumors SF3P is going to be renamed to 2nm so that is what probably happened. But the original plan was SF3E (2022), then SF3 (2024), then SF3P (2025), then SF2 (2025) and then SF2P (2026).
SF2 relative to SF3 is supposed to be 1.05x density, 8% better performance and 16% lower power, kind of underwhelming. Then SF2P adds backside power delivery (this was due in 2026 but they are now saying may be 2025). SF2P also relative to SF3 is 1.11x density, 13% higher performance and 20% lower power.
I am more interested in the volume of all these supposed deals. I always hear about wins here and there and at the end of the day, TSMC volume increases more that everyone else combined. how large is meta volume? Obviously qualcomm volume is huge but how large is the samsung portion.
My estimate for Samsung volume is that 50% of its "foundry" volume is for Samsung parts. 50% is external customer. is this about right?
