Optimized QCA SRAM for Efficient MAC Operations in AI Workloads

This paper presents the design and implementation of a complete Static Random Access Memory (SRAM) cell using Quantum-dot Cellular Automata (QCA) technology. The proposed SRAM is designed to store critical data required for Machine Learning (ML) algorithm training, with a focus on enhancing data transfer rates, particularly in Multiply-Accumulate (MAC) operations. We have successfully implemented and tested the Quantum-dot inverted cell, demonstrating its functionality and potential for acceleration. Furthermore, the implementation of a QCA buffer, exhibiting 90% similarity to conventional buffers, confirms the reliability and scalability of our design for data storage processes. Building upon these foundational components, we have successfully realized the complete QCA-based SRAM architecture. The implemented SRAM demonstrates effective data storage and retrieval, offering a promising solution for efficient data storage and processing in AI hardware systems. This work validates the feasibility and benefits of QCA-based SRAM for enhancing performance in demanding AI applications.