In a nutshell: TSMC has simply introduced its complete lineup of 3nm magnificence nodes launching over the following 3 years. Its new FinFlex tech will give chip designers much more flexibility to optimize every usual mobile for the specified energy intake, efficiency, and density.
TSMC has simply unveiled its complete N3 circle of relatives of procedure nodes. Chip designers similar to AMD, Apple, Nvidia, or even Intel will use those nodes over the following couple of years to manufacture their bleeding-edge chips.
The Taiwanese corporate has a complete of 5 other 3nm magnificence nodes. N3 will start high-volume production later this yr, with the primary chips anticipated to achieve consumers early subsequent yr. N3E will release later with efficiency and potency enhancements, upper yields, however fairly decreased common sense density.
Round 2024, TSMC will convey out N3P, which makes a speciality of efficiency enhancements. N3S, which wasn’t featured in TSMC’s roadmap, used to be best in brief discussed in dialog via SVP Kevin Zhang.
In any case, N3X will pop out a couple of yr later and make allowance for very excessive efficiency at upper voltages, with potency and prices taking the again seat. This manner is very similar to the 5nm magnificence N4X procedure beginning quantity production subsequent yr.
TSMC’s N3 and N3E nodes may also fortify the corporate’s new FinFlex tech. These days, chip designers have to select one library for every block inside of an SoC. With FinFlex, they may not have this limitation and can be capable of mix ‘n match other libraries inside of every block.
They are able to use 2-1 (double-gate single-fin) FinFETs in some portions to be able to scale back energy intake and die measurement (value) and go for 3-2 FinFETs in different spaces the place most efficiency is paramount. In the meantime, 2-2 FinFETs supply a stability of measurement, efficiency, and tool intake.
TSMC additionally discussed its upcoming N2 procedure node, which can use gate-all-around field-effect transistors (GAAFETs), with plans to begin quantity production in the second one part of 2025.
In comparison to N3E, it’s going to reportedly draw 25-30 p.c much less energy on the identical frequency and make allowance for 10-15 p.c extra efficiency with the similar energy intake and transistor rely. In the meantime, chip density will reportedly build up via over 10 p.c.