Nvidia Platform Boosting Advanced Semiconductor Manufacturing Speed

TSMC and Nvidia Collaborate to Enhance Semiconductor Production

In a significant development for the semiconductor industry, TSMC has partnered with Nvidia to implement the cuLitho platform, aimed at revolutionizing the computational lithography process crucial for manufacturing advanced chips.

Computational lithography is a complex and compute-intensive task that involves intricate computations in electromagnetic physics, photochemistry, and computational geometry. Traditionally, chip foundries rely on vast data centers to perform these computations, but the conventional systems are becoming increasingly inefficient, consuming enormous amounts of time and energy.

Nvidia's cuLitho platform, powered by 350 Nvidia H100 Tensor Core GPU systems, is poised to replace 40,000 CPU systems, significantly reducing production time, energy consumption, and space requirements in data centers. The collaboration between TSMC and Nvidia has yielded remarkable improvements in performance, throughput, cycle time, and power efficiency, according to TSMC CEO C.C. Wei.

Moreover, cuLitho incorporates generative AI capabilities that could potentially double the speed by enhancing optical proximity correction (OPC) - a critical process in lithography that ensures accuracy in transferring circuit patterns onto silicon wafers.

By accelerating computational lithography, each mask can now be produced more efficiently, leading to quicker development of new chip features and enabling calculations that were previously unfeasible. This innovative approach could also facilitate the adoption of techniques like inverse lithography, which have remained in the realm of theoretical research due to extensive computational requirements.

For nearly two decades, researchers have explored the possibilities of designing circuits by generating the desired final pattern on the wafer and then computing the optimal photomask pattern. However, the computational demands have been a barrier to practical implementation - a challenge that cuLitho aims to overcome, paving the way for the production of next-generation semiconductors.