NVIDIA (NVDA, Financial) has significantly advanced the speed of computational lithography, a crucial step in chip manufacturing, with its cuLitho platform. This technology allows NVIDIA and TSMC to enhance the speed of curvilinear and traditional Manhattan process flows by 45 times and nearly 60 times, respectively. NVIDIA's generative AI application algorithms further boost these speeds by twofold.
During the "NVIDIA AI Summit," NVIDIA highlighted its collaboration with TSMC in accelerated computing. The adoption of the cuLitho platform in TSMC's production processes enhances the development speed of next-generation chip technologies while reducing energy consumption.
cuLitho introduces accelerated computing to computational lithography, enabling TSMC to overcome the physical limits of current production processes. This platform enhances the precision needed for advanced semiconductor chips, particularly as chip sizes shrink to 3nm and below, which demands more precise and computation-intensive lithography.
NVIDIA has reduced the computational workload traditionally needing 40,000 CPU systems to just 500 NVIDIA DGX H100 systems. This shift boosts daily photomask output by 3-5 times and reduces power consumption by ninefold.
Historically, the OPC (Optical Proximity Correction) process in semiconductor lithography required extensive computational resources. NVIDIA has showcased how accelerated computing and AI technologies significantly transform OPC, making physical simulations more accurate and less resource-intensive.
Current process changes in wafer manufacturing often necessitate OPC modifications, causing computational bottlenecks. The advancements with cuLitho drastically reduce these limitations, quickening the creation of masks and shortening the development cycles for new technology nodes.
Twenty years ago, inverse lithography technology was theorized but computationally impractical. With cuLitho, leading foundries can now effectively implement such techniques, paving the way for the next generation of semiconductor innovations.
