Configure Multi-Instance GPU (MIG): Practice Questions — Administration (NVIDIA-Certified Professional: AI Operations)
Configure Multi-Instance GPU (MIG) — Practice Questions This set of practice questions is designed to help candidates prepare for the Configure...
Configure Multi-Instance GPU (MIG) — Practice Questions
This set of practice questions is designed to help candidates prepare for the Configure Multi-Instance GPU (MIG) section of the NVIDIA-Certified Professional: AI Operations exam. Each question includes four options, the correct answer, and a brief explanation to reinforce key concepts.
What is the primary benefit of enabling Multi-Instance GPU (MIG) on an NVIDIA A100 GPU?
- A) Increasing the GPU clock speed for single workloads
- B) Partitioning a single GPU into multiple isolated instances for concurrent workloads
- C) Reducing power consumption by disabling unused GPU cores
- D) Automatically optimizing GPU memory allocation for deep learning models
Correct Answer: B
Explanation: MIG allows a single physical GPU to be partitioned into multiple isolated GPU instances, enabling concurrent execution of multiple workloads with guaranteed resource isolation.
Which command-line tool is used to configure MIG instances on supported NVIDIA GPUs?
- A) nvidia-smi
- B) kubectl
- C) slurmctld
- D) runai
Correct Answer: A
Explanation: The nvidia-smi tool provides commands to enable, configure, and manage MIG instances on supported GPUs.
Before creating MIG instances, what is a necessary step to prepare the GPU?
- A) Disable the GPU driver
- B) Enable MIG mode on the GPU
- C) Restart the Kubernetes cluster
- D) Install Run:ai platform
Correct Answer: B
Explanation: MIG mode must be enabled on the GPU before creating and managing MIG instances.
How many MIG instances can be created on a single NVIDIA A100 40GB GPU at maximum?
- A) 2
- B) 4
- C) 7
- D) 8
Correct Answer: C
Explanation: The NVIDIA A100 40GB GPU supports up to 7 MIG instances, each with isolated compute and memory resources.
Which of the following is NOT a valid MIG instance profile?
- A) 1g.5gb
- B) 2g.10gb
- C) 3g.20gb
- D) 7g.40gb
Correct Answer: D
Explanation: Valid MIG profiles are named with the format ng.mb, where n is the number of GPU compute units and mb is the memory size in GB. There is no 7g.40gb profile; the maximum number of compute units per instance is 3g.
What is the effect of deleting a MIG instance on a GPU?
- A) It disables MIG mode on the GPU
- B) It frees the allocated resources for reuse
- C) It permanently reduces GPU memory
- D) It restarts the GPU driver
Correct Answer: B
Explanation: Deleting a MIG instance releases the allocated GPU resources, making them available for new instances or workloads.
Which Kubernetes feature is essential for scheduling workloads on specific MIG instances?
- A) Node affinity
- B) Persistent volumes
- C) Device plugins
- D) ConfigMaps
Correct Answer: C
Explanation: NVIDIA’s device plugin for Kubernetes enables the scheduler to recognize and assign workloads to specific MIG instances.
What is the recommended practice when configuring MIG instances for AI workloads?
- A) Create the largest possible MIG instance for all workloads
- B) Partition the GPU into multiple smaller instances to maximize utilization
- C) Disable MIG mode to maximize raw GPU performance
- D) Use MIG only for non-AI workloads
Correct Answer: B
Explanation: Partitioning the GPU into multiple smaller MIG instances allows concurrent AI workloads to run efficiently, improving overall GPU utilization.
More in this topic
Ready to test your knowledge?
Put what you've learned into practice with a quick quiz and track your progress.
Test your knowledge →