News & Events

Cryogenic Systems for Ultra-Low Energy and Enhanced Performance

Nurzhan Zhuldassov

Supervised by Eby Friedman

426 Computer Studies Building

The computing industry is confronting the fundamental physical limits of transistor scaling which has for decades driven exponential growth in computational power. As transistor dimensions shrink below three nanometers, issues such as excessive heat density and quantum tunneling effects have emerged, limiting further reductions in the critical transistor dimensions. These limitations necessitate the exploration of alternative technologies and methodologies to sustain the advancement of computing capabilities.

Cryogenic computing has emerged as a promising solution to these challenges. Operating electronic systems at cryogenic temperatures offers several fundamental benefits: increased carrier mobility in semiconductors, reduced thermal noise, negligible leakage currents, superconductive behavior, and reduced electrical resistance. As a result, computing systems are exhibiting higher operating frequencies, improved reliability, reduced noise, and lower power consumption. These improvements support advancements in a variety of important applications, such as quantum computers and cryogenic cloud computing systems.

Multiple aspects of cryogenic computing systems are explored in this dissertation to leverage these benefits. Key areas of research include cryogenic CMOS circuits; particularly, dynamic logic circuits operating at low temperatures and the optimization of cryogenic computing systems across multiple temperature zones. An important issue in cryogenic computing systems explored in this dissertation is to identify which technology should be used at which temperature to maximize performance and energy efficiency. Methodologies are proposed to explore the physical characteristics of transistors operating at cryogenic temperatures, cryogenic transistor models, and cryogenic modeling and synthesis of computing systems based on a novel application of graph theory. A graph representation of the system is constructed in the pro- posed methodology, where each edge represents an individual circuit characterized by a power consumption and propagation delay. The primary objective of the method- ology is to determine the most energy efficient operating temperature for each circuit while ensuring the total system delay satisfies a predetermined constraint.

The research described in this dissertation proposes novel methodologies for the design and optimization of cryogenic computing systems, offering a pathway to over- come existing limitations in computing. By addressing challenges related to circuits operating at low temperatures and novel methods for thermal optimization, this dissertation lays the groundwork for future developments in high performance, energy efficient computing technologies.