ISC20 Featured Demo: Running Multiple Workloads on a Single A100 GPU

The NVIDIA A100 Tensor Core GPU includes a groundbreaking feature called Multi-Instance GPU (MIG), which partitions the GPU into as many as seven instances, each with dedicated compute, memory, and bandwidth. 

This allows multiple users to run their workloads on the same GPU, maximizing per-GPU utilization and user productivity. 

This demo runs an HPC simulation, an AI inference, and a debugging session simultaneously on a single A100 GPU. In the past, this would have required 3 GPUs and can now be addressed with a single A100 with one more slice to spare.

The NVIDIA Tensor Core A100 GPU’s MIG feature allows system admins to maximize GPU utilization by running more workloads on the same GPU and gives users faster access to run their simulations.

Watch now> 

See All ISC Demos and News>

    Leave a Reply