NVIDIA has announced that TSMC, the world’s leading foundry, and Synopsys, a leader in silicon design solutions, are implementing NVIDIA’s computational lithography platform to accelerate manufacturing and push past the physical limits of the next generation of advanced semiconductor chips.
By integrating NVIDIA cuLitho with its software, manufacturing processes, and systems, TSMC and Synopsys aim to speed up chip manufacturing, eventually supporting NVIDIA’s next-generation Blackwell architecture GPUs.
“Computational lithography is a fundamental pillar in chip manufacturing,” said Jensen Huang, founder and CEO of NVIDIA. “Our collaboration with TSMC and Synopsys, through cuLitho, applies accelerated computing and generative AI to explore new frontiers in semiconductor scaling.”
NVIDIA has also introduced new generative AI algorithms that enhance cuLitho, a GPU-accelerated computational lithography library, significantly improving the semiconductor manufacturing process compared to current CPU-based methods.
Leaders in advanced semiconductors advance with the cuLitho platform
Computational lithography is the most computationally intensive workload in the semiconductor manufacturing process, consuming tens of billions of hours per year on CPUs. A typical set of masks for a chip could require over 30 million CPU hours, necessitating large data centers within semiconductor foundries. With accelerated computing, 350 NVIDIA H100 systems can now replace 40,000 CPU systems, speeding up production time while reducing costs, space, and energy consumption.
“Our collaboration with NVIDIA to integrate accelerated computing into TSMC’s workflow has resulted in significant performance improvements, dramatically enhancing performance, shortening cycle times, and reducing energy requirements,” said Dr. C.C. Wei, CEO of TSMC.
Since its introduction last year, cuLitho has allowed TSMC to explore new opportunities for innovative pattern technologies. In cuLitho tests on shared workflows, the companies achieved a 45 times acceleration in curvilinear flows and an improvement of nearly 60 times in more traditional Manhattan-style flows.
“Sassine Ghazi, President and CEO of Synopsys, said: “Our collaboration with TSMC and NVIDIA is critical to enabling scaling at the angstrom level as we pioneer advanced technologies to reduce delivery time by orders of magnitude through the power of accelerated computing.”
Revolutionary generative AI support for computational lithography
NVIDIA has developed algorithms to apply generative AI and further enhance the value of the cuLitho platform. The new generative AI workflow offers an additional 2x acceleration in processes accelerated by cuLitho. The application of generative AI allows for creating an almost perfect inverse mask to compensate for light diffraction. The final mask is then derived by traditional and physically rigorous methods, accelerating the overall optical proximity correction (OPC) process by a factor of two.
Changes in the current manufacturing process require a review of OPC, increasing the computational requirements and creating bottlenecks in the factory development cycle. These costs and bottlenecks are alleviated with accelerated computing provided by cuLitho and generative AI, allowing factories to allocate available computing capacity and engineering bandwidth to design more innovative solutions in the development of new technologies for 2nm and beyond.