Multi-Instance GPU (MIG) configuration for AI and HPC: Practice Questions — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)

Practice Questions: Multi-Instance GPU (MIG) Configuration for AI and HPC

This set of exam-style multiple-choice questions focuses on the Multi-Instance GPU (MIG) configuration aspect of the Physical Layer Management domain for the NVIDIA-Certified Professional: AI Infrastructure certification. These questions will help reinforce your understanding of configuring MIG for AI and HPC workloads.

1. What is the primary benefit of enabling Multi-Instance GPU (MIG) on an NVIDIA A100 GPU?

   • A) Increasing the GPU's clock speed for better performance
   • B) Partitioning a single GPU into multiple isolated instances for concurrent workloads
   • C) Reducing power consumption by disabling unused cores
   • D) Automatically updating GPU drivers for compatibility

   Correct answer: B

   Explanation: MIG allows a single physical GPU to be partitioned into multiple isolated GPU instances, enabling concurrent execution of multiple AI or HPC workloads with guaranteed resource isolation.

2. Which command-line tool is primarily used to configure and manage MIG instances on NVIDIA GPUs?

   • A) nvidia-smi
   • B) nvtop
   • C) nvidia-debugdump
   • D) nvidia-settings

   Correct answer: A

   Explanation: The nvidia-smi tool is the standard utility used to enable, configure, and monitor MIG instances on supported NVIDIA GPUs.

3. When configuring MIG, what is the maximum number of instances that can be created on a single NVIDIA A100 GPU?

   • A) 4
   • B) 7
   • C) 8
   • D) 16

   Correct answer: C

   Explanation: The NVIDIA A100 GPU supports up to 7 MIG instances, but the maximum number of instances depends on the specific GPU model and partition sizes. For the A100, up to 7 instances can be configured.

4. Which of the following is a valid reason to use MIG in HPC environments?

   • A) To increase single-thread CPU performance
   • B) To enable multiple users or jobs to share a single GPU securely and efficiently
   • C) To reduce network latency between nodes
   • D) To automatically balance CPU and GPU workloads

   Correct answer: B

   Explanation: MIG enables partitioning of GPU resources so multiple users or jobs can run isolated workloads on the same physical GPU, improving utilization and security in HPC environments.

5. Which of the following steps is required before creating MIG instances on an NVIDIA GPU?

   • A) Disabling the GPU's ECC memory
   • B) Enabling MIG mode on the GPU
   • C) Installing CUDA Toolkit version 9.0 or earlier
   • D) Updating the BIOS to the latest version

   Correct answer: B

   Explanation: MIG mode must be explicitly enabled on the GPU before any MIG instances can be created or managed.

6. How does MIG ensure workload isolation between different GPU instances?

   • A) By using separate physical GPUs for each instance
   • B) By partitioning GPU resources such as compute cores, memory, and cache into isolated slices
   • C) By running workloads sequentially instead of concurrently
   • D) By limiting GPU clock speeds for each instance

   Correct answer: B

   Explanation: MIG partitions the GPU’s hardware resources into isolated instances, each with dedicated compute cores, memory, and cache, ensuring no interference between workloads.

7. Which NVIDIA platform feature complements MIG for network and storage acceleration in AI infrastructure?

   • A) CUDA-X AI
   • B) BlueField Data Processing Unit (DPU)
   • C) TensorRT
   • D) NVLink

   Correct answer: B

   Explanation: The BlueField DPU platform complements MIG by offloading networking and storage tasks, improving overall AI infrastructure performance and management.

8. What is a common challenge when configuring MIG instances for mixed AI and HPC workloads?

   • A) Ensuring GPU instances have adequate and balanced resource allocation for diverse workload requirements
   • B) Installing multiple GPU drivers simultaneously
   • C) Disabling MIG mode to improve performance
   • D) Using only one MIG instance per GPU

   Correct answer: A

   Explanation: Different AI and HPC workloads may require varying GPU resources; careful planning and configuration of MIG instance sizes are necessary to optimize performance and resource utilization.

More in this topic

Multi-Instance GPU (MIG) configuration for AI and HPC: Worked Example — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)Multi-Instance GPU (MIG) configuration for AI and HPC: Quick Reference — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)Multi-Instance GPU (MIG) configuration for AI and HPC: Common Mistakes — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)Physical Layer Management — NVIDIA-Certified Professional: AI InfrastructureBlueField network platform configuration and management — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)Multi-Instance GPU (MIG) configuration for AI and HPC — Physical Layer Management (NVIDIA-Certified Professional: AI Infrastructure)