Server performance optimization: Practice Questions — Troubleshoot and Optimize (NVIDIA-Certified Professional: AI Infrastructure)
Server Performance Optimization – Practice Questions These practice questions focus on server performance optimization within advanced NVIDIA AI...
Server Performance Optimization – Practice Questions
These practice questions focus on server performance optimization within advanced NVIDIA AI infrastructure, a key topic for the NVIDIA-Certified Professional: AI Infrastructure exam. Each question includes four options, the correct answer, and an explanation to help reinforce your understanding.
Which of the following is the most effective first step when diagnosing a server experiencing degraded AI workload performance?
- A. Replace all storage drives
- B. Check GPU utilization and temperature metrics
- C. Reinstall the operating system
- D. Upgrade the server’s network interface card
Correct Answer: B
Explanation: Monitoring GPU utilization and temperature is critical because GPUs are the primary compute resource in AI workloads. Identifying bottlenecks or thermal throttling early helps target optimization efforts effectively.
What is the primary benefit of enabling NVIDIA Multi-Instance GPU (MIG) on a server?
- A. Increases CPU clock speed
- B. Allows partitioning of a single GPU into multiple isolated instances
- C. Automatically updates GPU drivers
- D. Reduces power consumption by disabling unused GPUs
Correct Answer: B
Explanation: MIG enables a single GPU to be partitioned into multiple instances, improving resource utilization and performance isolation for concurrent AI workloads.
Which storage optimization technique can reduce AI training bottlenecks caused by slow data access?
- A. Using RAID 0 (striping) to increase throughput
- B. Disabling write caching
- C. Using a single HDD instead of SSD
- D. Increasing CPU core count
Correct Answer: A
Explanation: RAID 0 stripes data across multiple drives, increasing read/write throughput, which is beneficial for high-throughput AI training data pipelines.
During server performance tuning, which metric indicates that the CPU is a bottleneck for AI inference workloads?
- A. GPU utilization consistently at 100%
- B. CPU utilization near 100% while GPU utilization is low
- C. Network latency below 1 ms
- D. Memory usage below 50%
Correct Answer: B
Explanation: High CPU utilization with low GPU usage suggests the CPU is limiting the workload, causing the GPU to wait and reducing overall performance.
Which NVIDIA tool is best suited for identifying performance bottlenecks in AI workloads running on GPUs?
- A. NVIDIA Nsight Systems
- B. NVIDIA Control Panel
- C. Windows Task Manager
- D. NVIDIA GeForce Experience
Correct Answer: A
Explanation: NVIDIA Nsight Systems provides detailed profiling and tracing of GPU workloads, helping to pinpoint performance bottlenecks in AI applications.
What is a common cause of server performance degradation related to memory in AI infrastructure?
- A. Insufficient GPU VRAM for model size
- B. Overclocked CPU cores
- C. Using RAID 5 for storage
- D. High network bandwidth
Correct Answer: A
Explanation: Insufficient GPU VRAM forces data swapping between GPU and system memory, significantly slowing AI model training and inference.
How can NUMA (Non-Uniform Memory Access) awareness improve server performance for AI workloads?
- A. By distributing GPU workloads evenly across all nodes
- B. By optimizing memory access patterns to reduce latency
- C. By disabling hyper-threading on CPUs
- D. By increasing network packet size
Correct Answer: B
Explanation: NUMA-aware scheduling ensures that CPU cores access local memory nodes, reducing latency and improving performance for memory-intensive AI workloads.
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 →