NVIDIA AI Infrastructure - NCP-AII FREE EXAM DUMPS QUESTIONS & ANSWERS]

Question 1

A user encounters "permission denied" errors when running GPU-accelerated containers on a Secure Boot-enabled system. What resolves this?

A. Run Docker with sudofor elevated privileges. B. Enroll the MOK and sign NVIDIA kernel modules. C. Reinstall Docker without the NVIDIA runtime. D. Disable SELinux to relax unnecessary security policies.

Discussion 0

Correct Answer: B Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 2

One of the nodes in a cluster is not running as fast as the others and the system administrator needs to check the status of the GPUs on that system. What command should be used?

A. nvidia-gpu-status B. lspci | grep NVIDIA C. iblinkinfo D. nvidia-smi

Discussion 0

Correct Answer: D Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 3

An engineer needs to verify NVLink isolation on a single node with 8 GPUs. Which NCCL test configuration stresses switch bisection bandwidth?

A. Use reduce_scatter_pref -b 8 -e 16G -f2 -g 4 B. Use NCCL_TESTS_SPLIT= "DIV 8" with point-to-point tests C. Use all_reduce_perf -b 8 -e 16G -f2 -g 8 without splits D. Use all_reduce_perf -b 8 -e 16G -f2 -g 8 with NCCL_TESTS_SPLIT= "AND 0x1"

Discussion 0

Correct Answer: D Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 4

After NCCL burn-in reports "transport retry count exceeded", which corrective action addresses the underlying fabric issue?

A. Inspect InfiniBand link quality metrics (BER, symbol errors) and replace faulty cables B. Increase NCCL_IB_TIMEOUT to tolerate longer latencies C. Reduce message size to decrease network utilization D. Switch from Ring to Tree algorithms via NCCL_ALGO=TREE

Discussion 0

Correct Answer: A Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 5

A customer has just completed the first boot of their DGX system and is prompted to create an administrative user. What is the correct approach for setting up this user to ensure secure BMC and GRUB access?

A. Use "sysadmin" as the username and a simple password for ease of management. B. Create separate usernames for BMC and GRUB to maximize flexibility. C. Skip the creation of a new user and retain the default admin account for BMC and GRUB access. D. Create a unique, strong, lower-case username and password that will be used for both BMC and GRUB access, avoiding default or weak credentials.

Discussion 0

Correct Answer: D Vote an answer

Question 6

An engineer is reimaging a DGX system in a large cluster. Which method ensures the most efficient and secure remote installation without physical access?

A. Skip ISO verification and directly flash the OS to the disk via SSH. B. Use apt-get to upgrade the OS without rebooting the system. C. Build a software image on Base Command Manager and then reimage. D. Create a USB drive with the ISO and manually boot from it on the DGX system.

Discussion 0

Correct Answer: C Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 7

An administrator needs to manually deploy the BlueField image on a target DPU. The administrator downloads the new image file and needs to flash it to the hardware. Which command should the administrator use?

A. dd if=/root/bf.image of=/dev/bf/ bs=4096k B. apt install doca-runtime C. bfb-install --rshim D. /opt/mellanox/mlnx-fw-updater/mlnx_fw_updater.pl

Discussion 0

Correct Answer: C Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 8

You are preparing a GPU cluster for distributed AI training. Before running workloads, you need to validate the cluster's hardware. Which of the following steps is the most effective?

A. Run single-node benchmarks to confirm GPU performance on individual nodes. B. Verify CPU performance across all nodes, as it indicates overall system stability. C. Use NCCL tests to validate GPU-to-CPU communication and run node health checks. D. Confirm that all nodes boot successfully into the operating system without errors.

Discussion 0

Correct Answer: A Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 9

Which of the following steps are essential components of a recommended DGX cluster installation procedure? (Choose two.)

A. Configure networking by validating all interfaces on each node - ensuring proper InfiniBand/Ethernet connectivity - prior to installing cluster software. B. Group nodes by function during initial setup and assign them to relevant categories in the cluster management tool. C. Complete application containerization, run distributed jobs, validate node health or storage availability. D. Install Slurm on the head node and then configure the compute nodes' default OS images.

Discussion 0

Correct Answer: A,B Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 10

A leaf switch shows "FW Version Mismatch" alerts for transceivers after cluster expansion. Which tool validates transceiver firmware against expected versions?

A. mlxconfig B. flint C. ethtool D. iblinkinfo

Discussion 0

Correct Answer: B Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 11

A system administrator needs to change the RAID level of DGX station and use the script included in the DGX software with no data on the array. What are the options the administrator can choose?

A. RAID 10 and RAID 5 B. RAID 1 and RAID 5 C. RAID 0 and RAID 1 D. RAID 0 and RAID 5

Discussion 0

Correct Answer: D Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 12

Which of the following tests should be used to check for the lowest possible latency between two nodes in a fabric?

A. ib_write_lat B. ib_read_bw C. iperf D. ib_read_lat

Discussion 0

Correct Answer: A Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 13

An enterprise builds a Kubernetes-based AI platform using NVIDIA GPU Operator. After adding several new GPU nodes, administrators notice that NVIDIA drivers, container runtime components, and monitoring agents are installed automatically without manual intervention.
Which capability of GPU Operator enables this behavior?

A. Lifecycle management of GPU software components through Kubernetes B. Automatic firmware flashing of GPUs C. Automatic BIOS configuration for every worker node D. Dynamic migration of virtual machines between GPU nodes

Discussion 0

Correct Answer: A Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).

Question 14

An administrator is configuring node categories in BCM for a DGX BasePOD cluster. They need to group all NVIDIA DGX H200 nodes under a dedicated category for GPU-accelerated workloads. Which approach aligns with NVIDIA's recommended BCM practices?

A. Assign nodes to the "login" category to simplify Slurm integration. B. Use the existing "dgxnodes" category without modification, as it is preconfigured for all DGX systems. C. Avoid categories and configure each DGX node individually via CLI. D. Create a new "dgx-h200" category, assign all DGX H200 nodes to it.

Discussion 0

Correct Answer: D Vote an answer

Explanation: Only visible for FreeCram members. You can sign-up / login (it's free).