Nvidia has highlighted in its latest research paper that it can use a combination of artificial intelligence (AI) techniques to find better ways to place big groups of transistors. Researchers used reinforcement learning and added a second layer of AI on top of it to get better results.
Bill Dally, Nvidia chief scientist, said the work is important because chip manufacturing improvements are slowing with per-transistor costs in new generations of chip manufacturing technology now higher than previous generations. He pointed out that companies are no longer getting an economy from that scaling. “To continue to move forward and to deliver more value to customers, we can’t get it from cheaper transistors. We have to get it by being more clever on the design.”
Dally had said in 2022 that the demand side of Nvidia research tries to drive demand for the company’s products by developing software systems and techniques that need GPUs to run well. He shared that Nvidia has three different graphics research groups because they are always pushing the state of the art in computer graphics. Dally said they have five different AI groups because using GPUs to run AI is a huge thing and getting bigger.
“It’s natural as an expert in AI that we would want to take that AI and use it to design better chips. We do this in a couple of different ways. The first and most obvious way is we can take existing computer-aided design tools that we have and incorporated AI.” Dally gave an example about a tool that takes a map of where power is used in their GPUs, and predicts how far the voltage grid drops. He said running this on a conventional CAD tool takes three hours.
