TSMC, the global leader in semiconductor manufacturing, is adopting NVIDIA’s cutting-edge computational lithography platform, cuLitho, to revolutionize chip production and push the boundaries of semiconductor technology.
Computational lithography, essential for printing circuit patterns onto silicon wafers, is one of the most complex and resource-heavy steps in chip manufacturing. It involves intricate calculations based on physics, photochemistry, and computational geometry. Traditionally, these processes require vast CPU-powered data centers, making them a significant bottleneck in the industry.
Now, with NVIDIA’s accelerated computing power, 350 H100 Tensor Core GPUs can handle the same workload as 40,000 CPU systems. This breakthrough slashes production times, reduces costs, and cuts energy consumption—offering a more efficient path to producing advanced chips.
According to TSMC CEO Dr. C.C. Wei, the integration of NVIDIA’s GPU-based computing has resulted in dramatic improvements in speed, performance, and energy use across their operations.
NVIDIA has also introduced generative AI algorithms to its cuLitho platform, doubling the speed of computational lithography. These AI-driven solutions help optimize the process of creating masks used in the manufacturing of chips, accelerating the optical proximity correction (OPC) process by 2x.
This shift marks a major leap in semiconductor manufacturing. Techniques like inverse lithography, which have been explored for decades but were impractical due to their computational demands, are now becoming feasible with cuLitho’s power. This innovation enables faster development of next-generation semiconductors that are more powerful and efficient.
Through this partnership, TSMC and NVIDIA are setting the stage for future advancements in chip technology, transforming the industry and helping to meet growing global demands for high-performance semiconductors.
