The process of designing and verifying chips like the modern processors, controllers and sensors in cars, for example, can take years and require millions of dollars of R&D. Major OEMs like Ford and General Motors can be hesitant to move legacy designs from older, capacity-constrained chip fabs, to new, more modern process nodes that have the much-needed spare capacity to handle big production runs.
Generally speaking, re-targeting old chip designs to a new modern manufacturing process nodes is both costly and time-consuming, but that’s where AI (Artificial Intelligence) could come to the rescue. To put it simply, the hyper-iterative nature of machine learning is well-suited for certain tasks in the semiconductor design process, especially the mundane grunt-work of optimization for performance, power and area (silicon real estate) of chip designs.
Chip design tools bellwether, Synopsys, is making great strides at harnessing machine learning in many areas of chip design, from place and route (otherwise known as floor-planning of circuits), to verification (proving out designs so they’re bug free before production) and even someday pulling-in the actual software workloads that will run on chips in an effort to help further optimize designs. The latest AI-fueled innovation Synopsys has on tap is, you guessed it, Silicon Remastering of legacy chips in newer, more modern chip fab processes.
Could AI-Powered Silicon Remastering Be A Solution To The Chip Shortage?