NVIDIA NCP-AII Questions & Answers

This option serves the scenario “An engineer compares two long HPL burn-in periods that use different matrix dimensions and CUDA libraries without documenting the difference”. The controlling explanation is that Performance trend measurements are meaningful only when benchmark workload and software are comparable. The present scenario is distinguished by “A completed benchmark can conceal thermal stress that reduces repeatability and service life”.

This alternative addresses the separate case “An HPL stress campaign produces high performance but occasionally fails validation residuals after several hours”. This case illustrates that HPL correctness is a separate quality gate from throughput even for repeated long-running tests. The deciding requirement here is “A completed benchmark can conceal thermal stress that reduces repeatability and service life”.

The option would apply to “A one-minute HPL run is used for a rack acceptance contract that requires proving sustained computation under design power”. This case illustrates that Short high-performance measurements cannot establish long-term component stability and cooling capacity. This question instead requires “A completed benchmark can conceal thermal stress that reduces repeatability and service life”.

The option would apply to “An NCCL burn-in is interrupted by a scheduled switch update and the report falsely labels the hardware rack as failing”. The controlling explanation is that Burn-in failure interpretation requires knowledge of external disruptions to avoid false hardware diagnoses. The present scenario is distinguished by “Intermittent hardware or congestion faults may emerge only during sustained multi-node communication”.

The related but distinct situation is “A long NCCL test passes only with two ranks while production jobs regularly use 64 GPUs distributed across multiple racks”. The key distinction is that Collective failure probability and network behavior change with participant count and communication topology. The deciding requirement here is “Intermittent hardware or congestion faults may emerge only during sustained multi-node communication”.

This option serves the scenario “A collective burn-in reports consistent success but one rack contributes very low bus bandwidth and extends global completion time”. This case illustrates that Synchronous collectives are constrained by slow participants and imbalanced network paths. This question instead requires “Intermittent hardware or congestion faults may emerge only during sustained multi-node communication”.

The option would apply to “A NeMo training burn-in resumes from a checkpoint but silently changes tokenizer and model configuration between runs”. The strongest acceptance result is that A checkpoint restart reproduces the expected training state and documented configuration values, not merely that a command executed. This question instead requires “A controlled change improves samples per second while GPU and fabric remain within expected conditions”.

The proposed validation does not isolate the reported fault. An effective closure test requires evidence that A controlled change improves samples per second while GPU and fabric remain within expected conditions.

That observation does not establish operating effectiveness of Profile and tune dataset preprocessing and CPU worker placement before upgrading GPU networking. The needed evidence is that A controlled change improves samples per second while GPU and fabric remain within expected conditions.

This option serves the scenario “ClusterKit reports a slow InfiniBand fabric but operators run it during a large unrelated backup and treat the result as hardware failure”. A completed remediation must demonstrate that A controlled retest correlates workload occupancy, NIC counters and fabric throughput to explain the variance. This question instead requires “A recorded ClusterKit campaign reports per-node and cross-node results with no unexplained outliers”.

This alternative addresses the separate case “A new cluster's ClusterKit run omits three unreachable nodes from the result table and the engineer marks the entire fleet as passed”. A completed remediation must demonstrate that The final report lists every expected node with a pass, fail or explicit approved exception. For the stated problem, “A recorded ClusterKit campaign reports per-node and cross-node results with no unexplained outliers” addresses the central condition.

The candidate evidence checks Separate test harness deployment failures from actual hardware performance measurements in another context. Here the specific acceptance result is A recorded ClusterKit campaign reports per-node and cross-node results with no unexplained outliers.

This option serves the scenario “After reboot, a previously assigned MIG layout disappears and the scheduler tries to place jobs on nonexistent device UUIDs”. This case illustrates that MIG instance configurations may need recreation after GPU or system resets depending on platform and driver. The present scenario is distinguished by “Logical isolation is ineffective if scheduling incorrectly shares the same device instance across untrusted jobs”.

This alternative addresses the separate case “An administrator enables MIG mode on a supported GPU but creates no GPU or compute instances and the workload sees no device”. This case illustrates that MIG mode enablement alone does not create addressable GPU compute resources for applications. The deciding requirement here is “Logical isolation is ineffective if scheduling incorrectly shares the same device instance across untrusted jobs”.

The related but distinct situation is “A new multi-tenant inference service assigns a whole GPU to every tiny model despite approved MIG-capable hardware”. This case illustrates that MIG partitions expose isolated compute and memory allocations that can improve utilization for compatible jobs. For the stated problem, “Logical isolation is ineffective if scheduling incorrectly shares the same device instance across untrusted jobs” addresses the central condition.

This alternative addresses the separate case “A dual-port NIC exposes only one working port after an optic swap and the operator treats the whole adapter as failed”. The primary exposure is that Replacing a good NIC for an optic issue adds downtime and can disturb PCIe and driver state. For the stated problem, “Software reinstalls can mask a disconnected or failed accelerator without restoring capacity” addresses the central condition.

The option would apply to “A high-speed network card vanishes from a node after a power cycle while other cards remain healthy”. The primary exposure is that Assuming a global switch fault can delay isolation of one defective network card. The present scenario is distinguished by “Software reinstalls can mask a disconnected or failed accelerator without restoring capacity”.

The related but distinct situation is “One PSU in a redundant GPU server shows a fault LED and the system continues running while the operator assumes redundancy is intact”. The direct failure mode is that A second PSU or power-feed failure can shut down the server unexpectedly. The deciding requirement here is “Software reinstalls can mask a disconnected or failed accelerator without restoring capacity”.

The alternative may occur in another situation (A GPU server reports the correct accelerator count during receiving, but its rack acceptance sheet has no serial numbers or recorded firmware versions), but the principal consequence here is that Relying on in-band access alone can trap administrators outside a failed cluster during initial deployment.

The related but distinct situation is “A contractor declares a new GPU cluster ready based only on successful PXE boot while half the server NICs lack documented fabric connections”. The primary exposure is that Accepting a partially commissioned cluster can strand distributed jobs when communication crosses missing links. For the stated problem, “Relying on in-band access alone can trap administrators outside a failed cluster during initial deployment” addresses the central condition.

This alternative addresses the separate case “An operations team uses a single successful training run as evidence that six racks meet the contractual availability target”. Without correction, An untested redundant component can fail during real operations despite an isolated training success. The deciding requirement here is “Relying on in-band access alone can trap administrators outside a failed cluster during initial deployment”.

The related but distinct situation is “An AI inference server performs correctly at one request per second but misses latency targets when all accelerators share the CPU and network paths”. The direct fix is Measure representative parallel inference with GPU, CPU, NUMA and fabric utilization during readiness tests. The deciding requirement here is “Compare MIG partition state and usable memory against workload resource requirements before scheduling”.

For this finding, Run a permission- and runtime-aware CUDA smoke workload using the intended production account is less direct than Compare MIG partition state and usable memory against workload resource requirements before scheduling, because Logical GPU partition settings change the resources available to workloads regardless of physical board count.

The option would apply to “A node passes short GPU tests but its application intermittently throttles after twenty minutes of sustained work”. The appropriate intervention is Use sustained workload-like power, thermals and memory testing rather than accepting an idle or brief check. The present scenario is distinguished by “Compare MIG partition state and usable memory against workload resource requirements before scheduling”.

This statement could explain another scenario: A fabric acceptance team sees all links up but does not check whether storage ports were accidentally wired to GPU collective switches. In the present case the more specific mechanism is that Dual-rail redundancy depends on correct physical peer placement, not just two connected NIC ports.

The option would apply to “A migration swaps two similarly labeled fabric cables and only one of four MPI ranks loses inter-node connectivity”. This case illustrates that Partial workload failures can reflect a wiring error affecting only selected endpoints. This question instead requires “Dual-rail redundancy depends on correct physical peer placement, not just two connected NIC ports”.

The related but distinct situation is “After replacing several DAC cables, a technician fails to check switch port breakout profiles and the same rack has mismatched lane mappings”. This case illustrates that Physical breakout assemblies and logical switch profiles must use compatible lane allocation. For the stated problem, “Dual-rail redundancy depends on correct physical peer placement, not just two connected NIC ports” addresses the central condition.

The option would apply to “A rack of fan-cooled AI nodes is installed with blanking panels removed and mixed front-to-rear airflow direction”. The decisive evidence is that Per-server inlet probes and GPU sensors remain within approved limits during a representative sustained job. The present scenario is distinguished by “A controlled PDU feed shutdown leaves the server operating with a documented remaining power budget”.

The candidate evidence checks Validate peak draw, power redundancy, breaker headroom and supported PSU feed layout before commissioning in another context. Here the specific acceptance result is A controlled PDU feed shutdown leaves the server operating with a documented remaining power budget.

This option serves the scenario “A GPU server draws 10 kW at full training load but the facilities acceptance test used only idle operating-system power”. The strongest acceptance result is that Facility instrumentation remains below limits during the expected simultaneous full-rack burn-in, not merely that a command executed. This question instead requires “A controlled PDU feed shutdown leaves the server operating with a documented remaining power budget”.

The option would apply to “A BMC firmware flash process fails halfway through and the technician immediately reboots every chassis in the rack”. The controlling explanation is that Firmware interruption can leave a device unable to initialize; uncontrolled bulk reboot may expand the incident. This question instead requires “Firmware operations require a safe and reproducible maintenance boundary distinct from active workload execution”.

The related but distinct situation is “An HGX server cluster has mixed GPU and BMC firmware revisions after a partial maintenance window, but the team runs burn-in as if versions match”. This case illustrates that Firmware components interact across GPUs, baseboards, NICs and management controllers and require supported combinations. For the stated problem, “Firmware operations require a safe and reproducible maintenance boundary distinct from active workload execution” addresses the central condition.

This option serves the scenario “An AI factory firmware update changes BMC inventory fields and a monitoring dashboard now reports healthy GPUs as missing”. The controlling explanation is that Firmware updates can change management interfaces and monitoring expectations without changing installed hardware. The present scenario is distinguished by “Firmware operations require a safe and reproducible maintenance boundary distinct from active workload execution”.