NCP-AII Premium File
- 170 Questions & Answers
- Last Update: Nov 6, 2026
At an AI infrastructure site, a redundant GPU server shows one failed PSU and runs normally at idle, but the site still counts it as fully redundant. Which evidence MOST directly shows that the specific blocker in this scenario has been corrected?
Both PSU units report healthy status and a permitted redundancy check confirms the expected power-feed capacity. This observation tests the stated power-supply redundancy failure correction rather than unrelated node health. The underlying principle is: A redundant design loses failure tolerance when one supply is unavailable, even if current workloads run.
Post-service BMC inventory and electrical status identify the correct unit with no loss of required power capacity. That is acceptance evidence for psu safe replacement tracking; it does not establish that power-supply redundancy failure has been corrected. The relevant observation is: Both PSU units report healthy status and a permitted redundancy check confirms the expected power-feed capacity.
The original GPU passes its local diagnostics while the repaired fabric link clears optical errors. That is acceptance evidence for gpu versus cable replacement decision; it does not establish that power-supply redundancy failure has been corrected. The relevant observation is: Both PSU units report healthy status and a permitted redundancy check confirms the expected power-feed capacity.
Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths. That is acceptance evidence for network adapter field replacement; it does not establish that power-supply redundancy failure has been corrected. The relevant observation is: Both PSU units report healthy status and a permitted redundancy check confirms the expected power-feed capacity.
At an AI infrastructure site, a redundant GPU server shows one failed PSU and runs normally at idle, but the site still counts it as fully redundant. Which technical principle BEST explains the specific failure or solution in this scenario?
A redundant design loses failure tolerance when one supply is unavailable, even if current workloads run. This is why the operational remedy in this scenario is: Confirm the faulty power supply with BMC and power telemetry, then replace it under supported service procedures. A direct demonstration is: Both PSU units report healthy status and a permitted redundancy check confirms the expected power-feed capacity.
Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility. That principle fits network adapter field replacement, but does not explain the power-supply redundancy failure failure mechanism. The applicable explanation is: A redundant design loses failure tolerance when one supply is unavailable, even if current workloads run.
Remote NUMA memory access can incur additional latency and consume inter-socket bandwidth. That principle fits numa cpu affinity tuning, but does not explain the power-supply redundancy failure failure mechanism. The applicable explanation is: A redundant design loses failure tolerance when one supply is unavailable, even if current workloads run.
A cooling failure can directly cause device thermal limits even when CUDA libraries are healthy. That principle fits fan failure under training load, but does not explain the power-supply redundancy failure failure mechanism. The applicable explanation is: A redundant design loses failure tolerance when one supply is unavailable, even if current workloads run.
At an AI infrastructure site, a redundant GPU server shows one failed PSU and runs normally at idle, but the site still counts it as fully redundant. If the described issue remains uncorrected, which consequence is the MOST direct operational risk?
A second power issue may abruptly shut down the node. This follows from the specific unresolved power-supply redundancy failure condition. An appropriate mitigation is: Confirm the faulty power supply with BMC and power telemetry, then replace it under supported service procedures.
Training nodes may idle on file opens even with fast accelerator and network hardware. That is a consequence of parallel filesystem metadata bottleneck, not the most direct risk of the stated power-supply redundancy failure defect. The scenario-specific exposure is: A second power issue may abruptly shut down the node.
Removing the wrong PSU could eliminate remaining power and abort the node. That is a consequence of psu safe replacement tracking, not the most direct risk of the stated power-supply redundancy failure defect. The scenario-specific exposure is: A second power issue may abruptly shut down the node.
Unverified replacements can create fabric partitioning or unexpected congestion. That is a consequence of network adapter field replacement, not the most direct risk of the stated power-supply redundancy failure defect. The scenario-specific exposure is: A second power issue may abruptly shut down the node.
At an AI infrastructure site, a field service team replaces a failed GPU but skips post-replacement burn-in because NVIDIA SMI displays the new serial number. Which operational response MOST directly resolves the specific blocker described before this equipment or service is accepted?
Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration. This addresses the reported accelerator hardware swap qualification blocker, because hardware presence proves identification, not sustained operational reliability after service. The acceptance evidence is: The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance.
Drain the affected node, correlate Xid codes with NVIDIA diagnostics and service guidance, and isolate the device before returning it to service. This is an appropriate response to gpu xid incident isolation, but the described incident is accelerator hardware swap qualification rather than that separate fault. It would not directly establish the required correction: Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration.
Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs. This is an appropriate response to network adapter field replacement, but the described incident is accelerator hardware swap qualification rather than that separate fault. It would not directly establish the required correction: Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration.
Replace the identified supported optical component after validating physical evidence; do not replace healthy GPUs. This is an appropriate response to gpu versus cable replacement decision, but the described incident is accelerator hardware swap qualification rather than that separate fault. It would not directly establish the required correction: Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration.
At an AI infrastructure site, a field service team replaces a failed GPU but skips post-replacement burn-in because NVIDIA SMI displays the new serial number. Which evidence MOST directly shows that the specific blocker in this scenario has been corrected?
The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance. This observation tests the stated accelerator hardware swap qualification correction rather than unrelated node health. The underlying principle is: Hardware presence proves identification, not sustained operational reliability after service.
A controlled preprocessing benchmark and memory-channel inventory reflect the intended platform configuration. That is acceptance evidence for cpu memory-channel imbalance; it does not establish that accelerator hardware swap qualification has been corrected. The relevant observation is: The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance.
Post-service BMC inventory and electrical status identify the correct unit with no loss of required power capacity. That is acceptance evidence for psu safe replacement tracking; it does not establish that accelerator hardware swap qualification has been corrected. The relevant observation is: The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance.
Follow-up diagnostics and representative jobs show no recurring Xid under the monitored acceptance workload. That is acceptance evidence for gpu xid incident isolation; it does not establish that accelerator hardware swap qualification has been corrected. The relevant observation is: The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance.
At an AI infrastructure site, a field service team replaces a failed GPU but skips post-replacement burn-in because NVIDIA SMI displays the new serial number. Which technical principle BEST explains the specific failure or solution in this scenario?
Hardware presence proves identification, not sustained operational reliability after service. This is why the operational remedy in this scenario is: Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration. A direct demonstration is: The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance.
Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility. That principle fits network adapter field replacement, but does not explain the accelerator hardware swap qualification failure mechanism. The applicable explanation is: Hardware presence proves identification, not sustained operational reliability after service.
GPU error codes can identify driver, memory, hardware, or communication fault classes requiring evidence-based triage. That principle fits gpu xid incident isolation, but does not explain the accelerator hardware swap qualification failure mechanism. The applicable explanation is: Hardware presence proves identification, not sustained operational reliability after service.
Layer-one or adapter faults can impair traffic even when upper-layer software still connects. That principle fits network card link errors, but does not explain the accelerator hardware swap qualification failure mechanism. The applicable explanation is: Hardware presence proves identification, not sustained operational reliability after service.
At an AI infrastructure site, a field service team replaces a failed GPU but skips post-replacement burn-in because NVIDIA SMI displays the new serial number. If the described issue remains uncorrected, which consequence is the MOST direct operational risk?
An untested replacement may fail in production and cause repeat outages. This follows from the specific unresolved accelerator hardware swap qualification condition. An appropriate mitigation is: Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration.
Unverified replacements can create fabric partitioning or unexpected congestion. That is a consequence of network adapter field replacement, not the most direct risk of the stated accelerator hardware swap qualification defect. The scenario-specific exposure is: An untested replacement may fail in production and cause repeat outages.
A second power issue may abruptly shut down the node. That is a consequence of power-supply redundancy failure, not the most direct risk of the stated accelerator hardware swap qualification defect. The scenario-specific exposure is: An untested replacement may fail in production and cause repeat outages.
Training nodes may idle on file opens even with fast accelerator and network hardware. That is a consequence of parallel filesystem metadata bottleneck, not the most direct risk of the stated accelerator hardware swap qualification defect. The scenario-specific exposure is: An untested replacement may fail in production and cause repeat outages.
At an AI infrastructure site, a failed ConnectX adapter is exchanged, and the server boots, but the new card has the wrong firmware baseline and network port mapping. Which operational response MOST directly resolves the specific blocker described before this equipment or service is accepted?
Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs. This addresses the reported network adapter field replacement blocker, because replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility. The acceptance evidence is: Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths.
Require physical safety checks, device diagnostics, link validation, and representative stress testing before reintegration. This is an appropriate response to accelerator hardware swap qualification, but the described incident is network adapter field replacement rather than that separate fault. It would not directly establish the required correction: Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs.
Replace the identified supported optical component after validating physical evidence; do not replace healthy GPUs. This is an appropriate response to gpu versus cable replacement decision, but the described incident is network adapter field replacement rather than that separate fault. It would not directly establish the required correction: Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs.
Inspect the physical network adapter, port, cable, and counters to localize the failing path before application tuning. This is an appropriate response to network card link errors, but the described incident is network adapter field replacement rather than that separate fault. It would not directly establish the required correction: Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs.
At an AI infrastructure site, a failed ConnectX adapter is exchanged, and the server boots, but the new card has the wrong firmware baseline and network port mapping. Which evidence MOST directly shows that the specific blocker in this scenario has been corrected?
Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths. This observation tests the stated network adapter field replacement correction rather than unrelated node health. The underlying principle is: Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility.
Fan sensors, temperatures, and sustained GPU clocks return to supported ranges after maintenance. That is acceptance evidence for fan failure under training load; it does not establish that network adapter field replacement has been corrected. The relevant observation is: Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths.
The replacement device is correctly inventoried and passes memory, fabric, and workload acceptance. That is acceptance evidence for accelerator hardware swap qualification; it does not establish that network adapter field replacement has been corrected. The relevant observation is: Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths.
Post-service BMC inventory and electrical status identify the correct unit with no loss of required power capacity. That is acceptance evidence for psu safe replacement tracking; it does not establish that network adapter field replacement has been corrected. The relevant observation is: Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths.
At an AI infrastructure site, a failed ConnectX adapter is exchanged, and the server boots, but the new card has the wrong firmware baseline and network port mapping. Which technical principle BEST explains the specific failure or solution in this scenario?
Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility. This is why the operational remedy in this scenario is: Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs. A direct demonstration is: Node and switch telemetry show the correct ports, supported firmware, and stable traffic through all required paths.
Hardware presence proves identification, not sustained operational reliability after service. That principle fits accelerator hardware swap qualification, but does not explain the network adapter field replacement failure mechanism. The applicable explanation is: Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility.
GPU error codes can identify driver, memory, hardware, or communication fault classes requiring evidence-based triage. That principle fits gpu xid incident isolation, but does not explain the network adapter field replacement failure mechanism. The applicable explanation is: Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility.
Concentrated I/O queues and uneven storage target placement create hotspots despite adequate aggregate capacity. That principle fits storage hotspot during checkpointing, but does not explain the network adapter field replacement failure mechanism. The applicable explanation is: Replacing a NIC changes physical identity and may require restoring configuration and firmware compatibility.
At an AI infrastructure site, a failed ConnectX adapter is exchanged, and the server boots, but the new card has the wrong firmware baseline and network port mapping. If the described issue remains uncorrected, which consequence is the MOST direct operational risk?
Unverified replacements can create fabric partitioning or unexpected congestion. This follows from the specific unresolved network adapter field replacement condition. An appropriate mitigation is: Validate replacement card identity, supported firmware, cable mapping, and intended fabric features before enabling jobs.
Removing the wrong PSU could eliminate remaining power and abort the node. That is a consequence of psu safe replacement tracking, not the most direct risk of the stated network adapter field replacement defect. The scenario-specific exposure is: Unverified replacements can create fabric partitioning or unexpected congestion.
Poor placement can reduce model training throughput even with healthy GPUs. That is a consequence of numa cpu affinity tuning, not the most direct risk of the stated network adapter field replacement defect. The scenario-specific exposure is: Unverified replacements can create fabric partitioning or unexpected congestion.
An untested replacement may fail in production and cause repeat outages. That is a consequence of accelerator hardware swap qualification, not the most direct risk of the stated network adapter field replacement defect. The scenario-specific exposure is: Unverified replacements can create fabric partitioning or unexpected congestion.
Please check your mailbox for a message from support@examlabs.com and follow the directions.