NVIDIA-Certified Associate AI Infrastructure and Operations (NCA-AIIO)

Exam practice questions

15 / 140 questions

Free preview · Answers and explanations includedPage 1 of 2

AI InfrastructureNCA-AIIO practice

A training job requires a measured peak device-memory allocation of 62 GB per GPU at its approved settings. The team will not change precision or partition the model. Which GPU characteristic must meet this requirement first?

Select one answerAPeak matrix throughput regardless of memory capacityBHost RAM capacity treated as interchangeable with device memoryCUsable device-memory capacity above the measured requirement with suitable headroomDThe number of GPU cores without checking allocations

AI InfrastructureNCA-AIIO practice

Host CPUs spend substantial time on infrastructure networking while application workloads need more CPU capacity. Which role of a BlueField DPU is relevant to evaluate?

Select one answerAIncreasing model Tensor Core throughput without changing infrastructure workBReplacing the training framework with a packet routerCUsing the DPU as additional transparent GPU framebuffer memoryDOffloading supported networking and infrastructure processing

AI OperationsNCA-AIIO practice

An operator investigates repeated computation failures on one GPU. Which telemetry is relevant to checking device health?

Select one answerAOnly average application accuracyBOnly the number of queued jobsCOnly peak nominal memory bandwidthDECC-related error indicators and diagnostic results where supported

Essential AI KnowledgeNCA-AIIO practice

An image-classification service must respond to 99% of customer requests within 150 milliseconds. Which measure should most directly guide its inference deployment tuning?

Select one answerAMean latency measured at negligible trafficBMaximum training samples processed per secondC99th-percentile end-to-end request latencyDPeak GPU utilization during an isolated benchmark

Essential AI KnowledgeNCA-AIIO practice

A developer must add a custom parallel calculation to an existing C++ application running on NVIDIA GPUs. Which platform provides the GPU programming model, compiler, and runtime needed for this work?

Select one answerACUDABTriton Inference ServerCDCGMDNVIDIA GPU Operator

Essential AI KnowledgeNCA-AIIO practice

A data team wants to accelerate dataframe preparation and supported machine-learning algorithms on GPUs while staying close to familiar Python data-science workflows. Which NVIDIA ecosystem should it evaluate?

Select one answerARivaBDCGMCRAPIDSDDeepStream

AI InfrastructureNCA-AIIO practice

A model’s working set cannot fit on any single available GPU. The serving framework supports partitioning the model across devices. Which scaling approach directly addresses the capacity problem?

Select one answerAAdditional data-parallel replicas of the full model on undersized devicesBAdditional serving processes on the same undersized GPUCModel parallelism across multiple GPUsDA larger host filesystem cache without model partitioning

Essential AI KnowledgeNCA-AIIO practice

An application applies the same independent calculation to millions of pixels. Which architectural feature makes a GPU a strong candidate for accelerating this portion of the workload?

Select one answerALow-latency serial execution as the dominant design goalBHigh-throughput execution of many parallel operationsCAutomatic speedup without expressing parallel workDElimination of memory-bandwidth constraints

Essential AI KnowledgeNCA-AIIO practice

A logistics company previously had only a few incomplete examples for demand prediction. It now has a large, representative, consistently labeled history. Which improvement most directly supports learning a useful predictive model?

Select one answerAIncreasing only GPU memory while retaining incomplete examplesBUsing a faster inference network without updating the training dataCChanging the serving batch size without improving examplesDAccess to more relevant, higher-quality training data

AI OperationsNCA-AIIO practice

A GPU server’s operating system has stopped responding. An authorized operator needs to inspect hardware status remotely before deciding on recovery. Which access path should be used?

Select one answerAAn SSH service running inside the unresponsive OSBA metrics agent that requires the failed OS to respondCThe BMC through the protected out-of-band management networkDAn inference health endpoint hosted on the failed process

AI InfrastructureNCA-AIIO practice

A synchronous distributed training job frequently exchanges large gradients among nodes. Which network characteristics are especially important to evaluate?

Select one answerAOnly single-node GPU memory capacityBOnly out-of-band management throughputCHigh usable bandwidth and low communication latencyDOnly archive-storage capacity without communication tests

Essential AI KnowledgeNCA-AIIO practice

A manufacturer has thousands of component photos labeled acceptable or defective. It wants a model to learn the relationship between each photo and its label. Which learning approach matches this task?

Select one answerAUnsupervised clusteringBSupervised learningCReinforcement learning from action rewardsDRule-based classification without fitting a model

AI InfrastructureNCA-AIIO practice

An organization’s policy requires a sensitive training dataset to remain within its own controlled facility. Which deployment choice most directly meets the stated location requirement?

Select one answerAUse an external region solely because it has the lowest priceBRun the workload in an appropriately controlled on-premises environmentCUse external cloud storage with encryption as a substitute for the location ruleDStage the full dataset externally whenever local GPUs are busy

AI InfrastructureNCA-AIIO practice

During one measurement period, a data center uses 1.2 MWh of total facility energy and its IT equipment uses 1.0 MWh. What is its power usage effectiveness (PUE) for that period?

Select one answerA1.2B0.2C0.83D2.2

AI OperationsNCA-AIIO practice

A research cluster receives batch training jobs from many users and must allocate nodes and GPUs according to queues and policies. Which software role is central to this requirement?

Select one answerAAn inference-only model serverBA GPU metrics exporterCA container image registryDA workload manager such as Slurm

Complete your exam preparation

Continue with the full question bank

Practise free with answers, explanations, and Community discussions. Join Community to unlock the complete question bank and timed mock exams.

Join Community Choose access for your certification

NVIDIA-Certified Associate AI Infrastructure and Operations (NCA-AIIO)

Exam practice questions

15 / 140 questions

Free preview · Answers and explanations includedPage 1 of 2

AI InfrastructureNCA-AIIO practice

A training job requires a measured peak device-memory allocation of 62 GB per GPU at its approved settings. The team will not change precision or partition the model. Which GPU characteristic must meet this requirement first?

AI InfrastructureNCA-AIIO practice

Host CPUs spend substantial time on infrastructure networking while application workloads need more CPU capacity. Which role of a BlueField DPU is relevant to evaluate?

AI OperationsNCA-AIIO practice

An operator investigates repeated computation failures on one GPU. Which telemetry is relevant to checking device health?

Select one answerAOnly average application accuracyBOnly the number of queued jobsCOnly peak nominal memory bandwidthDECC-related error indicators and diagnostic results where supported

Essential AI KnowledgeNCA-AIIO practice

An image-classification service must respond to 99% of customer requests within 150 milliseconds. Which measure should most directly guide its inference deployment tuning?

Essential AI KnowledgeNCA-AIIO practice

A developer must add a custom parallel calculation to an existing C++ application running on NVIDIA GPUs. Which platform provides the GPU programming model, compiler, and runtime needed for this work?

Select one answerACUDABTriton Inference ServerCDCGMDNVIDIA GPU Operator

Essential AI KnowledgeNCA-AIIO practice

A data team wants to accelerate dataframe preparation and supported machine-learning algorithms on GPUs while staying close to familiar Python data-science workflows. Which NVIDIA ecosystem should it evaluate?

Select one answerARivaBDCGMCRAPIDSDDeepStream

AI InfrastructureNCA-AIIO practice

A model’s working set cannot fit on any single available GPU. The serving framework supports partitioning the model across devices. Which scaling approach directly addresses the capacity problem?

Essential AI KnowledgeNCA-AIIO practice

An application applies the same independent calculation to millions of pixels. Which architectural feature makes a GPU a strong candidate for accelerating this portion of the workload?

Essential AI KnowledgeNCA-AIIO practice

A logistics company previously had only a few incomplete examples for demand prediction. It now has a large, representative, consistently labeled history. Which improvement most directly supports learning a useful predictive model?

AI OperationsNCA-AIIO practice

A GPU server’s operating system has stopped responding. An authorized operator needs to inspect hardware status remotely before deciding on recovery. Which access path should be used?

AI InfrastructureNCA-AIIO practice

A synchronous distributed training job frequently exchanges large gradients among nodes. Which network characteristics are especially important to evaluate?

Essential AI KnowledgeNCA-AIIO practice

A manufacturer has thousands of component photos labeled acceptable or defective. It wants a model to learn the relationship between each photo and its label. Which learning approach matches this task?

Select one answerAUnsupervised clusteringBSupervised learningCReinforcement learning from action rewardsDRule-based classification without fitting a model

AI InfrastructureNCA-AIIO practice

An organization’s policy requires a sensitive training dataset to remain within its own controlled facility. Which deployment choice most directly meets the stated location requirement?

AI InfrastructureNCA-AIIO practice

During one measurement period, a data center uses 1.2 MWh of total facility energy and its IT equipment uses 1.0 MWh. What is its power usage effectiveness (PUE) for that period?

Select one answerA1.2B0.2C0.83D2.2

AI OperationsNCA-AIIO practice

A research cluster receives batch training jobs from many users and must allocate nodes and GPUs according to queues and policies. Which software role is central to this requirement?

Select one answerAAn inference-only model serverBA GPU metrics exporterCA container image registryDA workload manager such as Slurm

Complete your exam preparation

Continue with the full question bank

Practise free with answers, explanations, and Community discussions. Join Community to unlock the complete question bank and timed mock exams.

Join Community Choose access for your certification