PROCESSOR INNOVATION AND AI PERFORMANCE

A decade ago, popular processing units were Central Processing Units (CPUs) and Graphics Processing Units (GPUs). Advances in artificial intelligence have skyrocketed the demand for specialised hardware. Along with GPUs, machine learning researchers have started using Accelerated Processing Unit (APUs), Tensor Processing Units (TPUs), and Neural Processing Units (NPUs). In short, the CPU manages overall tasks, the GPU handles graphics and visuals, the APU combines the functionality of the CPU and GPU, and the NPU is dedicated to accelerating machine learning and artificial intelligence tasks. A TPU on the other hand, is a specialised application-specific integrated circuit (ASIC) developed by Google for accelerating machine learning workloads. TPUs efficiently perform essential neural network tasks, such as matrix multiplications or other tensor operations.