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 89

Question 880

During an NVIDIA AI factory acceptance test, engineers discover: A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move. Which TWO complementary elements are needed: the correct technical action and the strongest independent diagnostic evidence? Select TWO answers.

  1. Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history.
  2. Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.
  3. Inspect DOM and EEPROM inventory, switch version, link logs, negotiated speed, FEC and retrain counters.
  4. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.
  5. Capture a reproducible preflight inventory of node hardware, software, connections and expected acceptance tests.

Explanation

Correct Answers

A:

This is the direct operational evidence to investigate: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. Observed scenario: A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives.

B:

This is an appropriate correction for InfiniBand link error and recovery counters: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. Observed scenario: A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives.

Incorrect Answers

C:

This alternative is relevant to Transceiver firmware and negotiated link test but does not directly resolve the current InfiniBand link error and recovery counters issue: Inspect DOM and EEPROM inventory, switch version, link logs, negotiated speed, FEC and retrain counters. Observed scenario: A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

E:

This alternative is relevant to Cluster-wide preflight hardware inventory but does not directly resolve the current InfiniBand link error and recovery counters issue: Capture a reproducible preflight inventory of node hardware, software, connections and expected acceptance tests. Observed scenario: A high-speed fabric stays connected but link recovery events and symbol errors increase after a cable move.

Question 881

A production GPU cluster review identifies this condition: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.
  2. Capture a reproducible preflight inventory of node hardware, software, connections and expected acceptance tests.
  3. Confirm InfiniBand physical and logical port states, subnet assignment and end-to-end reachability.
  4. Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

Explanation

Correct Answer

D:

This is an appropriate correction for InfiniBand link error and recovery counters: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

B:

This alternative is relevant to Cluster-wide preflight hardware inventory but does not directly resolve the current InfiniBand link error and recovery counters issue: Capture a reproducible preflight inventory of node hardware, software, connections and expected acceptance tests. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

C:

This alternative is relevant to InfiniBand link state LID and physical activation but does not directly resolve the current InfiniBand link error and recovery counters issue: Confirm InfiniBand physical and logical port states, subnet assignment and end-to-end reachability. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Question 882

In a professional AI infrastructure deployment, a technician observes: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. Which underlying infrastructure principle BEST explains the issue? Select ONE best answer.

  1. An active port can experience transient physical errors and recovery that degrade high-performance collectives.
  2. An enumerated HCA does not imply a fabric-ready port; InfiniBand state and LID assignment reflect additional setup.
  3. A representative performance benchmark is meaningful only when the tested hardware and software baseline are documented.
  4. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.

Explanation

Correct Answer

A:

This is the governing infrastructure principle: An active port can experience transient physical errors and recovery that degrade high-performance collectives. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Incorrect Answers

B:

That option concerns InfiniBand link state LID and physical activation, not the requirement in this case.

C:

That option concerns Cluster-wide preflight hardware inventory. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

Question 883

The cluster operations team needs to evaluate this situation: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. Which diagnostic evidence should be obtained FIRST to substantiate the likely fault or configuration gap? Select ONE best answer.

  1. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.
  2. Inspect ib_write_bw or ib_write_lat output, MTU, negotiated link speed and HCA or switch counters.
  3. Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes.
  4. Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history.

Explanation

Correct Answer

D:

This is the direct operational evidence to investigate: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

B:

That option concerns InfiniBand verbs microbenchmark bandwidth latency. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

C:

This alternative is relevant to CUDA device query and kernel smoke tests but does not directly resolve the current InfiniBand link error and recovery counters issue: Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Question 884

At a multi-node AI infrastructure handoff, the team records: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. After corrective work, which observable result BEST confirms that the requirement is met? Select ONE best answer.

  1. Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load.
  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. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.

Explanation

Correct Answer

A:

This is an observable verified result, not merely a claimed configuration change: Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

Incorrect Answers

B:

This alternative is relevant to InfiniBand link state LID and physical activation but does not directly resolve the current InfiniBand link error and recovery counters issue: All required HCA links are Active with valid fabric membership and can reach intended peer endpoints. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring.

C:

That option concerns InfiniBand subnet manager and route distribution. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. Observed scenario: NCCL bandwidth intermittently collapses although all switch ports report Up in five-minute monitoring. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. A direct check is: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. The acceptance condition is: Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load.

Question 885

During an NVIDIA AI factory acceptance test, engineers discover: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.
  2. Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior.
  3. Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.
  4. Check transceiver identity, firmware and supported FEC or link settings at both ends before stress testing.

Explanation

Correct Answer

C:

This is an appropriate correction for InfiniBand link error and recovery counters: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

B:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand link error and recovery counters issue: Run a controlled representative GPU stress with monitoring of temperatures, memory and power-limit behavior. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

D:

This alternative is relevant to Transceiver firmware and negotiated link test but does not directly resolve the current InfiniBand link error and recovery counters issue: Check transceiver identity, firmware and supported FEC or link settings at both ends before stress testing. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

Question 886

A production GPU cluster review identifies this condition: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. Which underlying infrastructure principle BEST explains the issue? Select ONE best answer.

  1. Sustained accelerator load can reveal thermal and electrical limits not visible during idle or brief smoke checks.
  2. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.
  3. An active port can experience transient physical errors and recovery that degrade high-performance collectives.
  4. HPL measures achieved dense linear-algebra throughput under a specific matrix and process mapping, not marketing peak clocks.

Explanation

Correct Answer

C:

This is the governing infrastructure principle: An active port can experience transient physical errors and recovery that degrade high-performance collectives. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

Incorrect Answers

A:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand link error and recovery counters issue: Sustained accelerator load can reveal thermal and electrical limits not visible during idle or brief smoke checks. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

B:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

D:

That option concerns HPL configuration matrix dimensions and performance. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

Question 887

In a professional AI infrastructure deployment, a technician observes: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. Which diagnostic evidence should be obtained FIRST to substantiate the likely fault or configuration gap? Select ONE best answer.

  1. Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes.
  2. Inspect GPU temperature and power traces, clock throttle reasons, error counters and completed stress duration.
  3. Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history.
  4. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.

Explanation

Correct Answer

C:

This is the direct operational evidence to investigate: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

Incorrect Answers

A:

This alternative is relevant to CUDA device query and kernel smoke tests but does not directly resolve the current InfiniBand link error and recovery counters issue: Inspect deviceQuery or equivalent CUDA sample output, driver-runtime versions and kernel launch error codes. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

B:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand link error and recovery counters issue: Inspect GPU temperature and power traces, clock throttle reasons, error counters and completed stress duration. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated.

Question 888

The cluster operations team needs to evaluate this situation: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. After corrective work, which observable result BEST confirms that the requirement is met? Select ONE best answer.

  1. The node sustains approved workload performance without unplanned throttling, errors or facility alarms.
  2. Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load.
  3. All targeted healthy GPUs complete the approved test suite or receive a documented fault disposition.
  4. Assume that any InfiniBand port in Active state must have zero physical faults or recovery events.

Explanation

Correct Answer

B:

This is an observable verified result, not merely a claimed configuration change: Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

Incorrect Answers

A:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current InfiniBand link error and recovery counters issue: The node sustains approved workload performance without unplanned throttling, errors or facility alarms. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path.

C:

That option concerns DCGM active single-node diagnostic suites. An active port can experience transient physical errors and recovery that degrade high-performance collectives.

D:

This tempting but unsafe shortcut is not a sufficient remediation for InfiniBand link error and recovery counters: Assume that any InfiniBand port in Active state must have zero physical faults or recovery events. Observed scenario: A technician clears error counters before capturing evidence and loses the pattern needed to isolate a flaky path. The governing mechanism is: An active port can experience transient physical errors and recovery that degrade high-performance collectives. The proportionate action is: Correlate cumulative and rate-based InfiniBand link counters with workloads and replace defective physical paths as indicated. A direct check is: Inspect symbol error, link recovery, port receive errors, timestamps, neighboring port counters and cable history. The acceptance condition is: Link errors stabilize below approved thresholds and NCCL performance recovers under repeated load.

Question 889

At a multi-node AI infrastructure handoff, the team records: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer. Which operational response is MOST appropriate to address the root issue? Select ONE best answer.

  1. Declare cable quality proven by the absence of hard link-down events while corrected FEC errors rise rapidly.
  2. Run a representative NeMo or equivalent distributed training smoke and burn-in covering data, communication and checkpoints.
  3. Collect optical power and FEC or BER data over representative load, then isolate unsafe optics or fibers.
  4. Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures.

Explanation

Correct Answer

C:

This is an appropriate correction for Optical FEC BER validation during burn-in: Collect optical power and FEC or BER data over representative load, then isolate unsafe optics or fibers. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

Incorrect Answers

A:

This tempting but unsafe shortcut is not a sufficient remediation for Optical FEC BER validation during burn-in: Declare cable quality proven by the absence of hard link-down events while corrected FEC errors rise rapidly.

B:

This alternative is relevant to NeMo distributed training end-to-end burn-in but does not directly resolve the current Optical FEC BER validation during burn-in issue: Run a representative NeMo or equivalent distributed training smoke and burn-in covering data, communication and checkpoints. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

D:

This alternative is relevant to DCGM active single-node diagnostic suites but does not directly resolve the current Optical FEC BER validation during burn-in issue: Schedule supported active DCGM diagnostics for the relevant GPU class and investigate detailed failures. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

Question 890

During an NVIDIA AI factory acceptance test, engineers discover: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer. Which TWO complementary elements are needed: the correct technical action and the strongest independent diagnostic evidence? Select TWO answers.

  1. Review transceiver diagnostics, FEC corrected and uncorrectable counters, BER trends and optical power.
  2. Execute supported CUDA runtime initialization and minimal kernel tests on each scheduled accelerator.
  3. Inspect GPU temperature and power traces, clock throttle reasons, error counters and completed stress duration.
  4. Collect optical power and FEC or BER data over representative load, then isolate unsafe optics or fibers.
  5. Declare cable quality proven by the absence of hard link-down events while corrected FEC errors rise rapidly.

Explanation

Correct Answers

A:

This is the direct operational evidence to investigate: Review transceiver diagnostics, FEC corrected and uncorrectable counters, BER trends and optical power. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

D:

This is an appropriate correction for Optical FEC BER validation during burn-in: Collect optical power and FEC or BER data over representative load, then isolate unsafe optics or fibers. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer. The governing mechanism is: Forward-error correction can mask deteriorating physical signal quality before an interface hard-down occurs.

Incorrect Answers

B:

This alternative is relevant to CUDA device query and kernel smoke tests but does not directly resolve the current Optical FEC BER validation during burn-in issue: Execute supported CUDA runtime initialization and minimal kernel tests on each scheduled accelerator. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

C:

This alternative is relevant to GPU thermal power and memory single-node burn-in but does not directly resolve the current Optical FEC BER validation during burn-in issue: Inspect GPU temperature and power traces, clock throttle reasons, error counters and completed stress duration. Observed scenario: A rack passes basic cable continuity but pre-FEC bit errors climb during full-rate data transfer.

E:

This tempting but unsafe shortcut is not a sufficient remediation for Optical FEC BER validation during burn-in: Declare cable quality proven by the absence of hard link-down events while corrected FEC errors rise rapidly. Forward-error correction can mask deteriorating physical signal quality before an interface hard-down occurs.

NCP-AI Infrastructure (NCP-AII) Practice Tests