TSMC aims to add a process developed for automotive driver-assistance systems to the portfolio offered by the fab the company aims to build in Arizona.
At the company’s annual symposium this week, CEO CC Wei said the process is a derivative of the N5 process, tweaked to “support the demand for AI-enabled ADAS. N5A is expected to be a long-lasting node. Our Arizona fab is to be one of the N5A production facilities”.
The US fab is expected to go into volume production in 2024; the N5A process is scheduled to move into production next year, with qualification to AEC-Q100 Grade 2 in the third quarter. Yuh Jier Mii, senior vice president of R&D, said the parent N5 process, which the foundry treats as a successor to the 7nm-class N7 node, moved into production quickly. “The speed of the N5 volume ramp is a new record [for the company],” he claimed.
Cheng Ming Lin, director of automotive business development, said N5A designs will be supported by Spice, PDKs and aging models, that cover aspects such as thermal-aware electromigration and end-of-life projections.
In addition to the automotive-focused N5A, TSMC is spinning off an RF-focused variant from its 7nm-class processes. Wei claimed it will offer “the PPA benefits of N6 while, tuned to RF solutions”.
According to Kevin Zhang, director of business development, N6RF can achieve peak frequencies of 350GHz, a 20 per cent increase over the prior RF process, which was based on a 16nm finFET technology. The shrink offers power reductions up to around 40 per cent.