AMD extends the usage of AI in chip designs

AMD has already started to integrate artificial intelligence in the design of processors, much as other significant chip manufacturers. In fact, Lisa Su, CEO of AMD, thinks that as the complexity of current CPUs rises exponentially, AI-enabled technologies may eventually take over chip design.

At the World Artificial Intelligence Conference (WAIC) in Shanghai in 2023, Su expressed her belief that AI will rule some aspects of chip design. She also emphasised the value of multidisciplinary cooperation for future improvements in hardware design, according to DigiTimes.

Chip design is an excellent application for AI, according to both Mark Papermaster, chief technology officer of AMD, and Jensen Huang, CEO of Nvidia. AMD already applies AI to designing, testing, and verifying semiconductors. Additionally, the business plans to use generative AI more extensively in the next semiconductor design applications.

According to Papermaster in May, AI is already being used by AMD in the chip design process, namely in the ‘place and route’ stage, where sub-blocks of chip designs are positioned and optimized for higher performance and reduced energy use. The process of developing an optimized layout is substantially accelerated by AI’s capacity to iterate continually and learn from patterns, enhancing performance and energy efficiency. According to Papermaster, AI will even spread to other crucial parts of chip design, such as microarchitecture designs, especially when some obstacles to IP protection are addressed.

AI is already used in verification suites to speed up the process of finding defects in a semiconductor as it develops, from the initial design to the verification and validation stages. AI also helps in the generation of test patterns. A chip design has billions of transistors, therefore thorough test coverage is necessary to ensure that the product is error-free when it leaves the production line. AI substantially speeds up the process and improves test coverage thanks to its capacity to learn from each subsequent run, spot coverage gaps, and change the testing strategy accordingly.

Chip designs are progressively being supported and helped by artificial intelligence (AI), which is becoming more and more important. All three of the top producers of electrical design automation (EDA) tools, Ansys, Cadence, and Synopsys, provide their customers with AI-enabled software, however Synopsys appears to be a step ahead of its rivals in this area.

The first end-to-end AI-driven EDA solution, Synopsys.ai, was introduced earlier this year by Synopsys. This makes it possible for designers to include AI into all phases of chip creation, including architecture, design, and manufacturing.