Dr. Paul Penzes, Vice President, Qualcomm, USA
K1-1: 9:00-09:45, Oct. 23, 2024, Wednesday
Title: Low-Power On-Device Computation for Future AI Expansion
Abstract:
Artificial intelligence (AI) has emerged as the primary driver of future semiconductor revenue growth with a projected ~$1 trillion market by 2030. The AI silicon market is dividing into a cloud-based high-performance market and a cost and power constrained edge market as vendors balance performance requirements, energy consumption, bandwidth, and privacy concerns. This presentation highlights the importance of low-power on-device ML computation in system-on-chip (SoC) processors at the edge and discusses methods to optimize power and performance considering manufacturing variation, dynamic workloads, and aging effects, including adaptive methods to respond to on die voltage variation and strategies to improve robustness and yield by design and EDA.
Biography:
Dr. Paul Penzes currently serves as Vice President of Engineering at Qualcomm, overseeing the Design Technology and Computer-Aided Design teams. Qualcomm stands at the forefront of the fabless industry, pioneering cutting-edge semiconductor technologies and advanced circuit techniques for high-volume chip manufacturing. Under Paul's leadership, his team spearheads research and development in low-power, high-speed circuits using cutting-edge process technologies, emphasizing scalability and robustness through advanced design automation. Prior to his role at Qualcomm, Paul contributed to the development of advanced design methodologies and circuit techniques at Broadcom (now Avago) as part of the Office of the CTO's team. Broadcom, renowned as a leading semiconductor design house, provided a fertile ground for Paul's diverse R&D expertise. Earlier, Paul served as chip lead at Myricom, a startup affiliated with Caltech, where he specialized in developing low-latency, high-speed interconnect chips. With over 20 years of experience in the semiconductor industry, Paul has made significant contributions to both research and development, resulting in numerous publications and patents. Paul holds a BS, MS, and PhD from the California Institute of Technology.