Abstract: As the limitations of traditional von Neumann architecture in handling big data and artificial intelligence applications become increasingly apparent, new computing architectures such as Computing-In-Memory (CIM) and neuromorphic computing have gradually become research hotspots. This paper reviews the development and research progress of neuromorphic devices and chips, introducing the principles of CIM and neuromorphic architectures. It focuses on the application of artificial synaptic devices, such as charge-based memory and resistive memory, in CIM and explores their potential in simulating biological neural networks. This paper also analyzes several typical neuromorphic chips, including IBM's TrueNorth, Stanford University's Neurogrid, and Intel's Loihi, demonstrating the prospects of neuromorphic computing in efficient and low-power computing. Through an in-depth discussion of these technologies, this paper provides new insights and directions for future neuromorphic research.