NVIDIA NCA-AIIO: NVIDIA-Certified Associate: AI Infrastructure and Operations

Last checked October 22, 2026 against the official NVIDIA exam page.

1. Question 1Essential AI Knowledge

   A team is choosing hardware for training deep neural networks. Which property of a GPU makes it better suited than a CPU for the large matrix operations at the core of training?

   1. AHigher single-thread clock speed that favours long sequential branching logic in code paths
   2. BMany cores that execute the same operation on many data elements in parallel
   3. CLarger per-core cache dedicated to operating system and kernel tasks
   4. DBuilt-in persistent NVMe storage for holding the training dataset
   Show answer and explanationHide answer

   Correct answer: B

   GPUs contain thousands of smaller cores designed for data-parallel arithmetic, which suits the matrix multiplications used in neural network training. CPUs are optimised for sequential, branchy workloads, and GPUs do not provide persistent storage.

2. Question 2AI Operations

   An administrator runs a Kubernetes cluster of GPU nodes and wants the driver, container runtime integration and device plugin to be installed and kept up to date automatically on each node. Which NVIDIA tool should they deploy?

   1. ANVIDIA GPU Operator
   2. BNVIDIA Triton Inference Server
   3. CNVIDIA RAPIDS
   4. DNVIDIA NeMo framework
   Show answer and explanationHide answer

   Correct answer: A

   The NVIDIA GPU Operator automates the lifecycle of the GPU software stack on Kubernetes nodes, including drivers, the container toolkit and the device plugin. Triton serves models, RAPIDS accelerates data science, and NeMo is a framework for building generative models.

3. Question 3AI Infrastructure

   A multi-GPU server must let its eight GPUs exchange data at very high bandwidth without going through the host CPU. Which interconnect provides this within the node?

   1. APCIe Gen3 slots without any switch
   2. BWi-Fi 6E adapters
   3. CUSB 3.2 ports
   4. DNVLink and NVSwitch
   Show answer and explanationHide answer

   Correct answer: D

   NVLink with NVSwitch provides high-bandwidth GPU-to-GPU links inside a system, avoiding the slower PCIe path through the CPU. PCIe is lower bandwidth, and wireless and USB connections are not used for GPU-to-GPU traffic.

4. Question 4Essential AI Knowledge

   Which description best matches inference as opposed to training in a deployed model?

   1. AUpdating model weights through backpropagation on new labelled data with gradient descent steps
   2. BCleaning and deduplicating a raw dataset before storage in the data lake with schema checks
   3. CLabelling a dataset by hand before the model is allowed to see it
   4. DRunning the forward pass with fixed weights to produce predictions for new inputs
   Show answer and explanationHide answer

   Correct answer: D

   Inference uses a trained model with frozen weights to compute outputs for new inputs, which needs only the forward pass. Backpropagation and weight updates happen in training, and labelling and cleaning are data preparation steps.

5. Question 5AI Operations

   An operator wants telemetry on GPU health such as temperature, power draw, clock speeds and utilization, collected from many servers for monitoring. Which NVIDIA component should be used to gather these metrics?

   1. ANVIDIA TensorRT-LLM Model Optimizer
   2. BDCGM (Data Center GPU Manager)
   3. CNVIDIA NGC Catalog Model Registry
   4. DNVIDIA cuDNN kernel library
   Show answer and explanationHide answer

   Correct answer: B

   DCGM collects and exports GPU health and telemetry data, and it integrates with monitoring systems such as Prometheus. TensorRT optimises inference engines, the NGC catalog hosts software, and cuDNN is a deep learning primitives library.