TSMC will start mass producing chips utilizing its modern N3 (3nm-class) manufacturing course of this September, based on a Business Occasions report that cites gear producers. The contract chipmaker will ship the primary merchandise made utilizing its N3 node to its prospects early subsequent yr.
Historically, TSMC begins its high-volume manufacturing (HVM) of a brand new node someday within the March to Might timeframe to provide sufficient chips for Apple’s newest iPhones, which generally launch in September. However the growth of TSMC’s N3 node took longer than normal, which is why Apple’s upcoming smartphone chips will use a special node. In distinction, the primary 3nm chips from TSMC will attain the HVM milestone solely in September, which is a bit later than what TSMC initially promised (a few months delay versus typical schedules). Nonetheless, the corporate will meet its objective of beginning N3 manufacturing within the second half of the yr.
|N3E vs N5||N3 vs N5|
|Pace Enchancment @ Identical Energy||+18%||+10% ~ 15%|
|Energy Discount @ Identical Pace||-34%||-25% ~ -30%|
|HVM Begin||Q2/Q3 2023||H2 2022|
When in comparison with the unique N5 manufacturing know-how, the preliminary N3 fabrication course of is projected to supply a ten% to fifteen% efficiency enchancment (on the identical energy and complexity), scale back energy consumption by 25% – 30% (on the identical pace and transistor depend), and improve logic density by round 1.6 occasions.
This unique N3 node has a slender course of window, which implies a lower-than-expected yield for some designs. TSMC is growing an N3E node with an improved course of window that may characteristic a barely decrease transistor density. That know-how is projected to enter mass manufacturing a few yr after N3, however there are indications that N3E will attain HVM sooner moderately than later. Ultimately, TSMC will add N3P, N3S, and N3X nodes to its 3nm household.
TSMC’s FinFlex tech is without doubt one of the key options of N3, enabling chip builders to combine and match completely different sorts of normal cells inside one block to precisely optimize efficiency, energy consumption, and space. FinFlex is especially useful for complicated issues like CPU or GPU cores, so corporations like Apple, AMD, Intel, and Nvidia will be capable of construct higher processors and graphics processors for PCs and high-performance computing purposes.
Apple is anticipated to be TSMC’s first buyer to undertake the N3 fabrication course of.