NVIDIA Computational Lithography Moves To TSMC’s Production

TSMC, a leading global semiconductor manufacturer, is using the NVIDIA Computational lithography platform to speed up production and push physics for next-generation semiconductor devices

Complex computations integrating photochemistry, computational geometry, iterative optimization, distributed computing, and electromagnetic physics are needed

The use of generative AI makes it possible to create an almost perfect inverse mask or inverse solution to take into consideration the light diffraction that occurs in computational lithography

This massive computational lithography speedup shortens the entire cycle time for building a new technology node by accelerating the production of each and every mask in the fab

NVIDIA cuLitho is a library that offers tools and algorithms that are designed to accelerate GPU-based computational lithography and semiconductor manufacturing processes by a significant margin

Computational lithography, a crucial process in the production of computer chips, is a sophisticated calculation combining distributed computing, photochemistry, computational geometry

A 40X performance speedup is powered by quicker inverse lithography technology (ILT), which produces correct photomasks more quickly

The job of 40,000 CPU systems may be completed by 500 NVIDIA Hopper GPU computers running cuLitho. That is 1/8 of the space and 1/9 of the power

Efficiency Two weeks may be used to process photomasks in a single night. It is possible to create 3X to 5X more masks each day