MIG
NVIDIA Multi-Instance GPU
NVIDIA introduced MIG(Multi-Instance GPU) since Ampere architecture.
MIG feature allows GPUs to be securely partitioned into up to seven separate GPU Instances for CUDA applications, so that multiple users use the same physical GPU with separate resources for optimal GPU utilization. [1]
MIG ensures to providing each instance's processors have separate and isolated paths through the entire memory system - the on-chip crossbar ports, L2 cache banks, memory controllers, and DRAM address busses are all assigned uniquely to an individual instanceenhanced isolation GPU resources.
With MIG, users will be able to see and schedule jobs on their new virtual GPU Instances as if they were physical GPUs. MIG works with Linux operating systems, supports containers using Docker Engine and hypervisors such as Red Hat Virtualization and VMware vSphere.