NCP-AII mock exam

Question 1, Cluster Test and VerificationAfter bring-up, you want a deep health test of every GPU in a node. What do you run?

• Anvidia-smi with no arguments, to see that each GPU is listed
• BA short PyTorch import test on each node
• Cibstat on every network port of the node
• DA DCGM diagnostic such as dcgmi diag

Question 2, Cluster Test and VerificationWhy run a sustained compute benchmark on a new cluster before handover?

• ATo confirm that the operating system image installed correctly
• BTo find faults that only appear under full load and heat
• CTo populate DNS records for each of the new nodes
• DTo create a baseline of user accounts and job quotas

Question 3, Cluster Test and VerificationA multi-node training test reports poor bandwidth between two nodes, while each node passes alone. What do you investigate?

• AThe firmware on the NVMe drives of those two nodes
• BThe GPU driver on a third node in the cluster
• CThe Python version in the test container
• DThe network fabric and links between those nodes

Question 4, System and Server Bring-UpBefore installing GPU drivers on a new server, which interface lets you manage it remotely, even when it is powered off?

• AA desktop remote app
• BA Python notebook
• CThe BMC
• DSSH on a data port

Question 5, System and Server Bring-Upnvidia-smi shows fewer GPUs than are installed. Which first step makes sense?

• AReplace the network switch and the cables to that node
• BChange the CUDA_VISIBLE_DEVICES value in the login shell
• CCheck the hardware inventory, firmware and PCIe status, then reseat if needed
• DReinstall the training framework and its CUDA libraries

Question 6, Control Plane Installation and ConfigurationWhat does a BlueField DPU offload from the host CPU?

• AGraphics rendering for remote desktops
• BNetworking, storage and security tasks
• CModel training math on large tensors
• DPython code execution for data loaders

Question 7, Troubleshooting and OptimizationOn an NVSwitch-based system, GPU-to-GPU communication fails after boot. Which service should you check?

• ANVIDIA Container Runtime
• BNVIDIA Fabric Manager
• CThe NVIDIA persistence daemon
• DThe Subnet Manager service

Question 8, Physical Layer ManagementA link keeps flapping and showing physical errors. What do you do first?

• ALower the link speed in the training configuration
• BInspect and reseat or replace the cable and transceiver
• CUpdate the training code to retry failed transfers
• DReinstall the OS on both connected nodes