Final yr, NVIDIA launched its cuLitho software program library, which guarantees to hurry up photomask growth by as much as 40 instances. Immediately, NVIDIA introduced a partnership with TSMC and Synopsys to implement its computational lithography platform for manufacturing use, and use the corporate’s next-generation Blackwell GPUs for AI and HPC functions.
The event of photomasks is an important step for each chip ever made, and NVIDIA’s cuLitho platform, enhanced with new generative AI algorithms, considerably hurries up this course of. NVIDIA says computational lithography consumes tens of billions of hours per yr on CPUs. By leveraging GPU-accelerated computational lithography, cuLitho considerably improves over conventional CPU-based strategies. For instance, 350 NVIDIA H100 techniques can now substitute 40,000 CPU techniques, leading to quicker manufacturing instances, decrease prices, and diminished house and energy necessities.
NVIDIA claims its new generative AI algorithms present a further 2x speedup on the already accelerated processes enabled by way of cuLitho. This enhancement is especially helpful for the optical proximity correction (OPC) course of, permitting the creation of near-perfect inverse masks to account for mild diffraction.
TSMC says that integrating cuLitho into its workflow has resulted in a 45x speedup of curvilinear flows and an nearly 60x enchancment in Manhattan-style flows. Curvilinear flows contain masks shapes represented by curves, whereas Manhattan masks shapes are restricted to horizontal or vertical orientations.
Synopsys, a number one developer of digital design automation (EDA), says that its Proteus masks synthesis software program operating on the NVIDIA cuLitho software program library has accelerated computational workloads in comparison with present CPU-based strategies. This acceleration is essential for enabling angstrom-level scaling and lowering turnaround time in chip manufacturing.
The collaboration between NVIDIA, TSMC, and Synopsys represents a major development in semiconductor manufacturing on the whole and cuLitho adoption specifically. By leveraging accelerated computing and generative AI, the companions are pushing semiconductor scaling potentialities and opening new innovation alternatives in chip designs.