NCP-AII Question 132
Single answerYou are tasked with configuring MIG (Multi-Instance GPU) on an NVIDIA A100 GPU to support multiple AI workloads simultaneously. The workloads require varying levels of compute resources. After enabling MIG mode on the GPU, what is the next step you need to perform to allocate GPU instances for these workloads?
- A
Use the NVIDIA System Management Interface (nvidia-smi) to create GPU instances with specific profiles.
- B
Restart the server to activate the MIG configuration.
- C
Install additional NVIDIA drivers to support MIG operations.
- D
Directly assign workload containers to the GPU without any further configuration.
Show answer and explanation
Correct answer: A
Explanation
To configure MIG on an NVIDIA GPU, you first enable MIG mode. After that, the critical next step is to use nvidia-smi to create GPU instances with specific profiles that allocate GPU resources for different workloads. This allows the GPU to be partitioned into multiple isolated instances, each with its own memory and compute resources, ensuring efficient resource utilization for AI and HPC tasks.
- A. Correct.
Correct. After enabling MIG mode, the next step is to use the NVIDIA System Management Interface (nvidia-smi) to create GPU instances with specific profiles that define the compute, memory, and other resources allocated to each instance.
- B. Incorrect.
Incorrect. While enabling MIG mode may require a restart, creating GPU instances is a separate step that does not involve restarting the server at this stage.
- C. Incorrect.
Incorrect. MIG functionality is supported by the installed NVIDIA drivers that are compatible with MIG-capable GPUs. Additional driver installation is not required once MIG support is already present.
- D. Incorrect.
Incorrect. Workload containers cannot be directly assigned to the GPU without first creating and configuring MIG instances to segment GPU resources appropriately.