GPU vs. CPU: A Comparative Study of Performance, Architecture, and NVIDIA's Innovations in Modern Computing

Graphics Processing Units (GPUs) and Central Processing Units (CPUs) are the foundations of contemporary computing in the rapidly developing field of computational technology. Their performance, architectural variations, and energy efficiency are compared in this research, emphasizing their distinct advantages across a range of workloads, including gaming, scientific computing, and artificial intelligence. This survey paper highlights how GPUs have transformed parallel processing with breakthroughs like Compute Unified Device Architecture (CUDA) cores and tensor operations. CPUs continue to play a crucial role in sequential and hybrid computing applications. Particular attention is given to NVIDIA's achievements, which include innovations in Turing and Ampere architectures, NVLink interconnects, and hybrid CPU-GPU systems such as the Grace Hopper superchip. The results highlight the complementary functions of these processors in heterogeneous systems and address their consequences for energy efficiency, High-Performance Computing (HPC), and hybrid architectures in the future.