Breaking bottlenecks: CPU optimization through architectural and neuromorphic techniques

This research explores two different approaches to improving how computers process information efficiently. The first part uses the Gem5 simulator to test and compare three types of CPU designs—Timing Simple CPU, Minor CPU, and O3CPU—by running a basic program. We looked at how features like pipelining, caching, and branch prediction affect how fast the program runs and how efficiently the CPU works. The second part focuses on recognizing handwritten digits from the MNIST dataset using two types of AI models. One model is a traditional neural network (MLP) that runs on a standard computer setup (Von Neumann architecture), and the other is a spiking neural network (SNN) that runs on a neuromorphic system, which mimics how the human brain works. Overall, this study shows how both architectural improvements and brain-inspired computing can help solve performance and efficiency issues in modern computing systems.