The Taiwan Semiconductor Research Institute (TSRI) on Tuesday presented an artificial intelligence (AI) system on chip (SoC) design and verification platform that it said can help academia reduce the time and cost needed to develop AI-related chips.
The platform, which has been in use for more than two years, is helping channel research and development teams to concentrate on the development of AI acceleration circuits, the key technology for developing AI chips, the TSRI said.
That is making it easier for academia to create SoCs for biomedical applications and verify their functions, according to the TSRI, which is part of the National Applied Research Laboratories (NARLabs).
TSRI associate research fellow Chang Yu-tsang said at a news conference that many technology companies have invested in R&D for AI chips, the TSRI said.
Apple, for example, has been developing its own chips since its release of the iPhone 8 to quickly unlock the iPhone and conserve power more efficiently, he said.
But such R&D requires top-end technology, and big investments in manpower, time and money to design and develop AI chips, creating a high threshold for entry into the AI chip design field.
The TSRI’s platform is lowering that threshold for academic researchers, Chang said. It uses Synopsys’ DesignWare ARC processor cores and provides an AI SoC design and verification environment.
Research teams at National Taiwan University, National Chiao Tung University National Chung Hsing University and National Sun Yat-sen University have been using the platform to develop chips for biomedical, autonomous vehicle and unmanned underwater vehicle applications, according to the TSRI.
Synopsys Taiwan President Robert Li said the keys to AI chip design include a standardized platform, customized applications and smart AI software.
The platform, which meets those goals, should accelerate designs of new chips by academic researchers and help Taiwan seize opportunities in the global AI applications market.