The most popular and comprehensive Open Source ECM platform
Let us know more
AI Chip Design: Moving from Scale Complexity to Systemic Complexity
A study by Google in 2021 found that AI can design advanced computer chips more effectively than engineers can do it using traditional chip-design software. This has become a common practice among companies designing chips now like Nvidia, Google, Samsung, and Synopsys.
Nvidia said that “machine-learning techniques such as deep convolutional neural networks and graph-based neural networks could eventually aid designer productivity and automate optimization tasks.”
Haoxing Mark Ren, Nvidia researcher, said that “you can design chips more efficiently, and it gives you the opportunity to explore more design space, which means you can make better chips.”
Mike Demler, analyst at the Linley Group, said that “artificial intelligence is well suited to arranging billions of transistors across a chip. It lends itself to these problems that have gotten massively complex. It will just become a standard part of the computational tool kit.”
Aart de Geus, chief executive of chip-design software maker Synopsys, said that “a machine will optimize everything, everything… Start with the specs of the chip, make some architectural decisions, and we automate all the rest.”
de Geus said that “we’re going to deliver 1,000 times performance at the end of the decade,” he said. “I have high confidence about that. I think we have had the end of classic Moore’s Law. And I believe this new era is moving from scale complexity to systemic complexity. I think the ambition is very high.”
