Deep Dive into IBM's new NanoStack 0.7nm Process Node for Chips – 666 MTr/mm2

4 hours ago

IBM announced a breakthrough in transistor technology called NanoStack, achieving a 0.7nm-class process node with 666 million transistors per square millimeter.
NanoStack uses a staggered sequential CFET (Complimentary FET) design, stacking NMOS and PMOS transistors vertically using a unique bonding technique to enable independent optimization and direct connectivity.
Key performance improvements include a 50% logic area scaling, 50% performance at iso-power, 70% efficiency at iso-performance, 40% SRAM scaling, and potential transistor densities exceeding 500 MTr/mm².
IBM's approach utilizes two different silicon crystal orientations ((001) for NMOS and (110) for PMOS) and addresses thermal and signal integrity challenges through advanced bonding and process innovations.
High-NA EUV lithography is beneficial but not strictly required; IBM plans to use it at Albany NanoTech to reduce patterning steps and improve cost and yield.
The technology is expected to take around five years to reach the market, with initial adoption likely in smartphones or small AI chiplets.
IBM emphasizes the need for EDA (Electronic Design Automation) tools to evolve for 3D transistor-level design, incorporating mechanical and thermal modeling alongside traditional PPA (Power, Performance, Area) considerations.

Hasty Briefsbeta