On March 19th, at the GTC 2024 conference, NVIDIA officially announced that TSMC (Taiwan Semiconductor Manufacturing Company) and Synopsys, Inc. will be the first to utilize the NVIDIA computational lithography platform in their production processes to expedite the manufacturing speed of next-generation advanced semiconductor chips and overcome physical limitations.
As known, TSMC is a global leader in semiconductor foundries, while Synopsys is a leading provider of chip-to-system design solutions. Both have integrated the NVIDIA cuLitho accelerated computational lithography platform into their software, manufacturing processes, and systems. This not only accelerates chip manufacturing speed but also speeds up support for the latest generation NVIDIA Blackwell architecture GPUs in the future.
In modern chip manufacturing processes, computational lithography is a crucial step and the most demanding workload in semiconductor manufacturing, requiring large-scale data centers. Over time, the evolution of silicon miniaturization has exponentially increased the demand for computation.
If CPUs were used for computation, hundreds of billions of hours would be consumed annually on computational lithography. For instance, a typical chip mask requires 30 million hours or more of CPU computation time. With accelerated computing, 350 NVIDIA H100 GPUs can now replace 40,000 CPU systems, reducing production time while also lowering costs, space, and power consumption.
Reportedly, NVIDIA's computational lithography platform can accelerate the most intensive computational workloads in semiconductor manufacturing by 40-60 times.
NVIDIA has also introduced a new generative AI algorithm that will further enhance the efficiency of cuLitho, greatly improving semiconductor manufacturing processes compared to current CPU-based methods.
"Computational lithography is the cornerstone of chip manufacturing," said NVIDIA founder and CEO Jensen Huang. "Through our collaboration with TSMC and Synopsys on cuLitho, applying accelerated computing and generative artificial intelligence, we have opened up new frontiers for semiconductor expansion."
