Scale GPU infrastructure for different use cases: Practice Questions — AI Infrastructure (NVIDIA-Certified Associate: AI Infrastructure and Operations)

Practice Questions: Scale GPU Infrastructure for Different Use Cases These multiple-choice questions are designed to help candidates prepare for the...

Practice Questions: Scale GPU Infrastructure for Different Use Cases

These multiple-choice questions are designed to help candidates prepare for the NVIDIA-Certified Associate: AI Infrastructure and Operations exam, focusing on scaling GPU infrastructure for various AI workloads.

Which factor is most critical when scaling GPU infrastructure for large-scale AI training?
- A. Number of CPU cores per server
- B. GPU memory capacity and interconnect bandwidth
- C. Size of the hard disk drive
- D. Number of network switches
Correct answer: B
Explanation: Large-scale AI training demands high GPU memory and fast inter-GPU communication (e.g., NVLink) to efficiently handle large models and datasets.
When scaling GPU infrastructure for inference workloads with low latency requirements, which approach is most appropriate?
- A. Use fewer GPUs with higher compute power
- B. Scale out horizontally with multiple smaller GPU nodes
- C. Increase CPU resources instead of GPUs
- D. Use cloud storage instead of local storage
Correct answer: B
Explanation: Horizontal scaling with multiple GPU nodes allows parallel processing of inference requests, reducing latency.
What is a key consideration when scaling GPU infrastructure for distributed training across multiple nodes?
- A. Ensuring sufficient PCIe lanes on each node
- B. Minimizing network latency and maximizing bandwidth between nodes
- C. Increasing the number of CPU threads
- D. Using slower storage devices to reduce cost
Correct answer: B
Explanation: Distributed training requires fast, low-latency inter-node communication (e.g., InfiniBand or high-speed Ethernet) to synchronize model updates efficiently.
Which GPU scaling strategy is best suited for AI workloads with highly variable demand?
- A. Fixed on-premises GPU cluster with static capacity
- B. Cloud-based GPU infrastructure with elastic scaling
- C. Single high-end GPU server
- D. Using CPUs instead of GPUs
Correct answer: B
Explanation: Cloud infrastructure enables dynamic scaling to match workload demand, optimizing cost and performance.
In a multi-tenant AI infrastructure, how can GPU resources be efficiently scaled and shared?
- A. Assign GPUs exclusively to each tenant without sharing
- B. Use GPU virtualization and container orchestration
- C. Limit GPU usage to batch jobs only
- D. Use only CPU resources for multi-tenant workloads
Correct answer: B
Explanation: GPU virtualization and container orchestration (e.g., Kubernetes with NVIDIA GPU support) allow flexible, efficient sharing and scaling of GPU resources among tenants.
What role does NVSwitch play when scaling GPU infrastructure?
- A. It provides power management for GPUs
- B. It enables high-bandwidth, low-latency GPU-to-GPU communication within a server
- C. It manages cooling for GPU clusters
- D. It connects GPUs to CPUs via PCIe
Correct answer: B
Explanation: NVSwitch is a high-speed switch enabling direct GPU-to-GPU communication, critical for scaling performance in multi-GPU servers.
Which of the following is a common bottleneck when scaling GPU infrastructure without adequate networking?
- A. Insufficient GPU memory
- B. Network congestion and latency
- C. CPU overheating
- D. Disk I/O speed
Correct answer: B
Explanation: Poor network performance can limit data transfer rates between GPUs and nodes, reducing overall training or inference efficiency.
When planning to scale GPU infrastructure for mixed AI workloads (training and inference), what is a best practice?
- A. Use separate dedicated GPU clusters optimized for each workload type
- B. Use a single GPU cluster with fixed configuration
- C. Prioritize CPU resources over GPUs
- D. Use only cloud infrastructure
Correct answer: A
Explanation: Training and inference have different performance and latency requirements; dedicated clusters allow optimized resource allocation.

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 →