In recent years, artificial intelligence-based solutions have disrupted the domain-centric computing architecture. The recent trends and growth in semiconductor technology, especially in system-on-chip (SOC), chiplet, memory, and interconnect (on-chip, die-to-die and chip-to-chip), build better disaggregated compute infrastructure with in-built, domain-specific accelerators (DSAs) at the silicon level. Also, various cognitive computing applications (example: speech recognition, face detection, natural language processing (NLP), and suspicious activity monitoring, etc.) are powered with deep neural networks (DNNs) using silicon.T o handle the massive computing requirement and bandwidth, off-chip computing will be moved to on-chip computing. The bottleneck battle between computing resource and data storage can be better addressed by artificial intelligence (AI) and domain-specific accelerators. Download the whitepaper to continue reading.
