This deep dive explores how the semiconductor industry transitioned from the golden age of Moore’s Law and Dennard Scaling to today’s complex 3D FinFET technology. The discussion reveals how traditional flat transistor scaling hit fundamental physical limits, forcing engineers to build vertical ‘fin’ structures that wrap gates around channels in three dimensions. However, this shift introduced new challenges where mechanical stress from neighboring components significantly affects transistor performance, making chip layout an active part of device physics rather than just a blueprint. The conversation examines how local layout effects from diffusion breaks and gate cuts can cause up to 12% performance variations, particularly impacting PMOS transistors. Using data from 30,000 test devices on actual 7-nanometer chips, the analysis shows how these microscopic mechanical forces directly impact modern AI supercomputers like NVIDIA’s Hopper and Blackwell systems, where billions of transistors must work in perfect coordination for optimal performance.
You can listen to all the podcasts by clicking here.