Siemens said on October 14 that integrating Nvidia’s cuLitho library into its Calibre manufacturing software has accelerated computational lithography workflows by as much as twelve times.
Computational lithography is the simulation step that predicts how a mask pattern will actually print on a wafer. At leading-edge nodes the number of shapes and corrections has grown to the point where runtime, not physics, sets the schedule.
cuLitho is a CUDA-X library that moves those calculations onto GPUs. Siemens has paired it with Calibre, the industry reference for mask synthesis, optical proximity correction and verification.
The companies framed the result as a way to keep advanced patterning tractable as AI chips drive larger die sizes and tighter pitches. Shorter simulation cycles give fabs more iterations to converge on a manufacturable mask.
Nvidia has invested heavily in positioning GPUs as fab infrastructure rather than only datacenter hardware. Earlier the same week it disclosed expanded GPU libraries for semiconductor manufacturing at SEMICON West.
