Graphics Processing Unit (GPU) is a specialized processor meant to accelerate graphics rendering or processing complex mathematical calculations such as during machine learning training.
GPUs improve the efficiency of parallel processing and are ideal for processing through large datasets during machine learning training. GPUs are offered in virtual and physical machines to provide additional computation power. If needed, GPUs are programmable so users can have more control on computational capabilities.
A multi-GPU is more than one GPU in a single computer system which increases processing power and accelerates parallel computing processing. With multiple GPUs, the workload is distributed across the GPUs.
A virtual GPU (vGPU) allows multiple virtual machines (VM) to share a physical GPU. vGPU utilizes and manages GPU resources.