Organizer and Moderator: Ali Keshavarzi, Stanford University
Co-Organizer: Ali Khakifirooz, Intel Corporation
Panelists: Tsu-Jae King Liu (UC Berkeley), Tom Lee (Stanford), Mark Lundstrom (Purdue), Mark Papermaster (AMD), Arijit Raychowdhury (Georgia Tech), Kevin Zhang (TSMC)
Before the IEDM meeting was in place, there was a lack of scalability of mechanical switches and vacuum tubes in rapidly growing telephony applications at the time that led to the need for a solid-state switch and the development of semiconductor transistor devices. The transistor is over 75 years old. The critical scaling of transistors has enabled integrated circuits with billions of interconnected transistors that are now fueling computing and their corresponding memory storage across various applications in the semiconductor industry. Moore’s Law and Dennard’s Scaling drove the era-defining computing that has spanned across a wide range of form factors and use cases from servers in the cloud, to desktops, and mobile devices at the edge. Today computing is at the core of AI applications with a diversity of computing processors and a range of computing power.
The demand for computing is on the rise with the continuing advancement of AI. The AI algorithms have become more complex requiring higher computing performance. As a result, computing power has increased on an unsustainable curve and uneconomical outcome. Should this trend continue, the golden age of AI will be short lived. Computing and its foundations are the chokepoint. The resulting path forward in the pursuit of energy-efficient computing requires a hard engineering focus and hence receives attention at IEDM and other IEEE conferences and technical communities.
How did we arrive at today’s problem? What is the path forward? What are the approaches for our unquenchable appetite for computing? Bringing efficiency to an exponential challenge? All these questions and more will be discussed during the special 70th anniversary IEDM 2024 evening panel.