cert
cert-1
cert-2

Pass NVIDIA NCP-AII Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers to Pass the Actual Exam!
30 Days Free Updates, Instant Download!

cert-5
cert-6
NCP-AII Exam - Verified By Experts
NCP-AII Premium File

NCP-AII Premium File

$89.99
$98.99
  • Premium File 170 Questions & Answers. Last Update: Nov 06, 2026

Whats Included:

  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
 
$98.99
$89.99
accept 14 downloads in the last 7 days
block-screenshots
NVIDIA NCP-AII Practice Test 87

Question 860

During an NVIDIA AI factory acceptance test, engineers discover: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage. Which TWO complementary elements are needed: the correct technical action and the strongest independent diagnostic evidence? Select TWO answers.

  1. Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures.
  2. Configure a supported active subnet manager and validated standby behavior with consistent fabric policy.
  3. Inspect HPL output for residual tests, return codes, numerical anomalies and rerun consistency.
  4. Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information.
  5. Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub.

Explanation

Correct Answers

B:

This is an appropriate correction for InfiniBand subnet manager and route distribution: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy. Observed scenario: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage. The governing mechanism is: InfiniBand requires subnet management for address assignment and routing across the switched fabric.

D:

This is the direct operational evidence to investigate: Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information. Observed scenario: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage. The governing mechanism is: InfiniBand requires subnet management for address assignment and routing across the switched fabric.

Incorrect Answers

A:

This alternative is relevant to DCGM active single-node diagnostic suites but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures. Observed scenario: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage.

C:

This alternative is relevant to HPL numerical result and residual validation but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Inspect HPL output for residual tests, return codes, numerical anomalies and rerun consistency. Observed scenario: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage.

E:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand subnet manager and route distribution: Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub. Observed scenario: A fabric has conflicting subnet managers and route/path behavior becomes unstable after a controller outage. The governing mechanism is: InfiniBand requires subnet management for address assignment and routing across the switched fabric. The proportionate action is: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy. A direct check is: Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information. The acceptance condition is: New and existing ports become correctly addressed and end-to-end fabric paths work after manager failover.

Question 861

A production GPU cluster review identifies this condition: A new node links electrically but cannot communicate with peers until fabric management discovers it. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures.
  2. Configure a supported active subnet manager and validated standby behavior with consistent fabric policy.
  3. Select a representative supported HPL problem size and process layout, then report achieved and reference performance honestly.
  4. Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub.

Explanation

Correct Answer

B:

This is an appropriate correction for InfiniBand subnet manager and route distribution: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Incorrect Answers

A:

This alternative is relevant to DCGM active single-node diagnostic suites but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

C:

This alternative is relevant to HPL configuration matrix dimensions and performance but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Select a representative supported HPL problem size and process layout, then report achieved and reference performance honestly. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand subnet manager and route distribution: Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub. A new node links electrically but cannot communicate with peers until fabric management discovers it.

Question 862

In a professional AI infrastructure deployment, a technician observes: A new node links electrically but cannot communicate with peers until fabric management discovers it. Which underlying infrastructure principle BEST explains the issue? Select ONE best answer.

  1. A representative performance benchmark is meaningful only when the tested hardware and software baseline are documented.
  2. InfiniBand requires subnet management for address assignment and routing across the switched fabric.
  3. Sustained accelerator load can reveal thermal and electrical limits not visible during idle or brief smoke checks.
  4. Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub.

Explanation

Correct Answer

B:

This is the governing infrastructure principle: InfiniBand requires subnet management for address assignment and routing across the switched fabric. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Incorrect Answers

A:

This alternative is relevant to Cluster-wide preflight hardware inventory but does not directly resolve the current InfiniBand subnet manager and route distribution issue: A representative performance benchmark is meaningful only when the tested hardware and software baseline are documented. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

C:

That option concerns GPU thermal power and memory single-node burn-in. The proportionate action is: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand subnet manager and route distribution: Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub. A new node links electrically but cannot communicate with peers until fabric management discovers it. InfiniBand requires subnet management for address assignment and routing across the switched fabric.

Question 863

The cluster operations team needs to evaluate this situation: A new node links electrically but cannot communicate with peers until fabric management discovers it. Which diagnostic evidence should be obtained FIRST to substantiate the likely fault or configuration gap? Select ONE best answer.

  1. Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub.
  2. Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information.
  3. Inspect HPL output for residual tests, return codes, numerical anomalies and rerun consistency.
  4. Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history.

Explanation

Correct Answer

B:

This is the direct operational evidence to investigate: Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand subnet manager and route distribution: Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it. The governing mechanism is: InfiniBand requires subnet management for address assignment and routing across the switched fabric. The proportionate action is: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy.

C:

This alternative is relevant to HPL numerical result and residual validation but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Inspect HPL output for residual tests, return codes, numerical anomalies and rerun consistency. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

D:

This alternative is relevant to InfiniBand link error and recovery counters but does not directly resolve the current InfiniBand subnet manager and route distribution issue: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Question 864

At a multi-node AI infrastructure handoff, the team records: A new node links electrically but cannot communicate with peers until fabric management discovers it. After corrective work, which observable result BEST confirms that the requirement is met? Select ONE best answer.

  1. Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub.
  2. All required HCA links are Active with valid fabric membership and can reach intended peer endpoints.
  3. New and existing ports become correctly addressed and end-to-end fabric paths work after manager failover.
  4. The node sustains approved workload performance without unplanned throttling, errors or facility alarms.

Explanation

Correct Answer

C:

This is an observable verified result, not merely a claimed configuration change: New and existing ports become correctly addressed and end-to-end fabric paths work after manager failover. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand subnet manager and route distribution: Assume unmanaged InfiniBand switches independently assign all LIDs and routes just like an ordinary Ethernet hub. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it. The governing mechanism is: InfiniBand requires subnet management for address assignment and routing across the switched fabric. The proportionate action is: Configure a supported active subnet manager and validated standby behavior with consistent fabric policy. A direct check is: Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information. The acceptance condition is: New and existing ports become correctly addressed and end-to-end fabric paths work after manager failover.

B:

This alternative is relevant to InfiniBand link state LID and physical activation but does not directly resolve the current InfiniBand subnet manager and route distribution issue: All required HCA links are Active with valid fabric membership and can reach intended peer endpoints. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

D:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand subnet manager and route distribution issue: The node sustains approved workload performance without unplanned throttling, errors or facility alarms. Observed scenario: A new node links electrically but cannot communicate with peers until fabric management discovers it.

Question 865

During an NVIDIA AI factory acceptance test, engineers discover: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.
  2. Run supported verbs bandwidth and latency tests across representative peers while examining physical counters.
  3. Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior.
  4. Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

Explanation

Correct Answer

B:

This is an appropriate correction for InfiniBand verbs microbenchmark bandwidth latency: Run supported verbs bandwidth and latency tests across representative peers while examining physical counters. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.

C:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

D:

This alternative is relevant to InfiniBand link error and recovery counters but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Question 866

A production GPU cluster review identifies this condition: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes. Which underlying infrastructure principle BEST explains the issue? Select ONE best answer.

  1. DCGM diagnostic suites exercise hardware and software health more deeply than passive status queries, sometimes consuming substantial resources.
  2. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.
  3. RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.
  4. A representative performance benchmark is meaningful only when the tested hardware and software baseline are documented.

Explanation

Correct Answer

C:

This is the governing infrastructure principle: RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Incorrect Answers

A:

This alternative is relevant to DCGM active single-node diagnostic suites but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: DCGM diagnostic suites exercise hardware and software health more deeply than passive status queries, sometimes consuming substantial resources. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

B:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency. RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.

D:

That option concerns Cluster-wide preflight hardware inventory. The proportionate action is: Run supported verbs bandwidth and latency tests across representative peers while examining physical counters.

Question 867

In a professional AI infrastructure deployment, a technician observes: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes. Which diagnostic evidence should be obtained FIRST to substantiate the likely fault or configuration gap? Select ONE best answer.

  1. Inspect ib_write_bw or ib_write_lat output, MTU, negotiated link speed and HCA or switch counters.
  2. Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes.
  3. Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information.
  4. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.

Explanation

Correct Answer

A:

This is the direct operational evidence to investigate: Inspect ib_write_bw or ib_write_lat output, MTU, negotiated link speed and HCA or switch counters. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Incorrect Answers

B:

This alternative is relevant to CUDA device query and kernel smoke tests but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

C:

This alternative is relevant to InfiniBand subnet manager and route distribution but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Inspect subnet-manager state, LIDs, partition policy, discovered topology and routing information. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency. A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Question 868

The cluster operations team needs to evaluate this situation: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes. After corrective work, which observable result BEST confirms that the requirement is met? Select ONE best answer.

  1. Representative peer pairs meet approved RDMA bandwidth and latency targets repeatably.
  2. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.
  3. The node sustains approved workload performance without unplanned throttling, errors or facility alarms.
  4. Representative workers sustain required read and write rates without unacceptable training stalls or checkpoint delays.

Explanation

Correct Answer

A:

This is an observable verified result, not merely a claimed configuration change: Representative peer pairs meet approved RDMA bandwidth and latency targets repeatably. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Incorrect Answers

B:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency. A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes. RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.

C:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: The node sustains approved workload performance without unplanned throttling, errors or facility alarms. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

D:

This alternative is relevant to Storage streaming bandwidth under GPU load but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Representative workers sustain required read and write rates without unacceptable training stalls or checkpoint delays. Observed scenario: A NIC line rate looks correct but small-message RDMA latency spikes between neighboring compute nodes.

Question 869

At a multi-node AI infrastructure handoff, the team records: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Test NCCL throughput and latency across representative message sizes and node counts with repeatable conditions.
  2. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.
  3. Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior.
  4. Run supported verbs bandwidth and latency tests across representative peers while examining physical counters.

Explanation

Correct Answer

D:

This is an appropriate correction for InfiniBand verbs microbenchmark bandwidth latency: Run supported verbs bandwidth and latency tests across representative peers while examining physical counters. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

Incorrect Answers

A:

That option concerns NCCL bandwidth and latency payload sweep. RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.

B:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency. An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

C:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

Question 870

During an NVIDIA AI factory acceptance test, engineers discover: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path. Which TWO complementary elements are needed: the correct technical action and the strongest independent diagnostic evidence? Select TWO answers.

  1. Inspect ib_write_bw or ib_write_lat output, MTU, negotiated link speed and HCA or switch counters.
  2. Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures.
  3. Run supported verbs bandwidth and latency tests across representative peers while examining physical counters.
  4. Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history.
  5. Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency.

Explanation

Correct Answers

A:

This is the direct operational evidence to investigate: Inspect ib_write_bw or ib_write_lat output, MTU, negotiated link speed and HCA or switch counters. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

C:

This is an appropriate correction for InfiniBand verbs microbenchmark bandwidth latency: Run supported verbs bandwidth and latency tests across representative peers while examining physical counters. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path. The governing mechanism is: RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.

Incorrect Answers

B:

This alternative is relevant to DCGM active single-node diagnostic suites but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

D:

This alternative is relevant to InfiniBand link error and recovery counters but does not directly resolve the current InfiniBand verbs microbenchmark bandwidth latency issue: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. Observed scenario: An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path.

E:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand verbs microbenchmark bandwidth latency: Treat a successful host TCP ping test as equivalent to verified InfiniBand verbs throughput and latency. An engineer relies on TCP iperf alone to certify the fabric's InfiniBand RDMA path. RDMA verbs benchmarks reveal transport behavior not fully measured by ordinary TCP connectivity tools.

NCP-AI Infrastructure (NCP-AII) Practice Tests