Graphics processing unit

A GPU is used to provide the high-throughput computation resources required for deep learning tasks such as model training, model validation, and model inference..

GPU requirements
Trained models are deployed to a graphical processing unit (GPU).
GPU storage
The GPU usage status bar on the menu shows the number of GPUs that are available.
Supported GPU devices
The following table lists the supported GPU devices for Maximo Visual Inspection and Maximo Visual Inspection Edge.
GPU splitting with NVIDIA Multi-Instance GPU
To optimize your GPU resources, you can use the NVIDIA Multi-Instance GPU (MIG) feature to partition a GPU into multiple separate GPUs.
Deployment of one model onto multiple GPUs
To optimize the inferencing speed of high-resolution models, you can deploy one model onto multiple GPUs.
CUDA compute capability 8.x for NVIDIA Ampere architecture GPUs
Several model types in Maximo Visual Inspection support CUDA compute capability 8.x for NVIDIA Ampere architecture GPUs.

Parent topic: Planning for Maximo Visual Inspection

Next topic: Storage requirements