Microsoft, investing in the OpenAI startup that founded ChatGPT, is actively implementing relevant developments in various areas, and Synopsys, with the support of this software giant, was able to develop a generative artificial intelligence system that helps engineers quickly design and search for new processors for errors in it.
Synopsys, a company that develops software tools for chip development, is talking about this reported This week he noted that Microsoft’s own processors for the Cobalt 100 and Maia 100 Azure cloud systems were also built with a customized version of the Copilot assistant. The Synopsys specialists who trained this system were faced with the task of achieving a minimum of errors in the operation of this assistant, since in the field of processor development the presence of errors is very expensive for companies. They typically invest years and hundreds of millions of dollars developing a single processor and the work of thousands of engineers. All design decisions must achieve 99.9% accuracy.
Another challenge in developing modern processors that contain billions of tiny transistors is finding possible errors and debugging the design. The customized version of Copilot copes with this task and speeds up the search for problem areas. A generative intelligence system in the early stages of design allows engineers to formulate technical specifications in strictly defined terms similar to a programming language. In the future, artificial intelligence will offer some standard design solutions, taking into account the experience of Synopsys specialists accumulated over decades of work. This assistant runs on the computing power of Microsoft Azure OpenAI. Perhaps after further testing, such development tools will also be offered to other Synopsys customers, not just Microsoft.