Certain workloads require more single threaded performance than is available in conventional CMOS CPUs. The performance of these CPUs is often limited by the large amounts of heat they produce. Many have discovered that by removing this heat and drastically increasing the clock frequency, they can obtain large boosts in performance.
Despite this, details of CMOS performance at extremely cold process corners are not well known. This project seeks to explore the performance of CMOS technology at extremely low temperatures from a more rigorous engineering perspective. The ultimate goal of this project is to determine the feasibility of engineering a CPU expressly for low temperature, high performance operation.
The project will proceed in several phases. The first involves gaining familiarization with sub-zero CPU cooling systems. This phase will also include an exploration of the performance of off-the-shelf CMOS CPUs in sub-zero conditions.
The second phase will involve using these observations to design and manufacture a prototype CMOS device, designed expressly to answer questions raised in the first phase, and to provide more precise insights into the performance of specific functional blocks. A platform for this chip, consisting of a motherboard and test interface, must also be created.
The third phase will consist of testing this new chip at extremely low temperatures. Each functional block will be stringently tested to determine what problems, if any, are solved by the cold, and what problems, if any, may arise.