NCP-AIN practice questions
NVIDIA · NCP-AIN · 200 questions
Original practice questions for the NVIDIA-Certified Professional: AI Networking (NCP-AIN) exam, covering AI data center design and optimization, NVIDIA Spectrum networking (Spectrum-X, RoCE, QoS/ECN/PFC, BGP-EVPN, NetQ, WJH, DOCA, SuperNIC), NVIDIA InfiniBand networking (provisioning, HA, PKeys, QoS, adaptive routing, UFM), Kubernetes Network Operator integration, troubleshooting tools, and NVUE/Ansible automation — framed in civic and municipal scenarios such as city GPU fabrics, public-health inference networks, transit and water ML clusters, and emergency-management AI fabrics.
This course contains the use of artificial intelligence.
About the NCP-AIN exam
- Exam fee
- $400 USD
- Time allowed
- 2 hours
- Questions
- 70-75
- Passing score
- Pass/Fail
- Languages
- English
Exam details published by the vendor, checked 18 October 2026. Vendors change fees and formats without notice — confirm on the vendor's own page before you book.
Practice Quizzes
Test your knowledge with standard 20-question practice sets.
Quiz 1
Quiz 2
Quiz 3
Quiz 4
Quiz 5
Quiz 6
Quiz 7
Quiz 8
Quiz 9
Quiz 10
Browse by Domain
Study specific topics at your own pace.
AI Data Center Design and Optimization · 10 questions
- A city GPU fabric planning board is designing a new AI factory row with GPU compute nodes and Spectrum leaf switches. Which role do NVIDIA BlueField DPUs play in that architecture?
- A public-health inference network expands with new scalable units. Which set best names the networking building blocks that glue the AI factory together?
- A transit authority AI factory diagram lists only GPUs and leaf-spine switches. What component should network engineers add before procurement to complete the AI factory networking picture?
- A county assessor imagery cluster is sized as scalable units. How should the architect explain switch-tier attachment for those units in the municipal AI factory?
- A city GPU fabric for multi-node training stalls on east-west congestion. Designers compare a rail-optimized topology to a classic Clos-only leaf-spine. Which choice best supports high-performance AI workloads?
- A municipal AI factory wants each GPU rail to land on the same switch set so NCCL rings avoid unnecessary hops. What does that rail-optimized goal achieve?
- Emergency-management model training will span eight nodes. Before cutover, what characteristic should planners verify to confirm the proposed topology is rail-optimized?
- A school-district IT GPU lab asks when GPU-to-GPU traffic should stay on NVLink/NVSwitch versus traverse the Spectrum or InfiniBand fabric. Which guidance is correct?
- A public-works pavement-scan training job shows poor all-reduce times. Engineers review GPU-to-GPU communication paths in the pod. What should they evaluate first?
- A housing authority multi-GPU node fails collective performance. How should the team decide whether the issue is GPU-to-GPU interconnect versus east-west fabric?
NVIDIA Spectrum Networking · 60 questions
- A public-health inference network needs clinic models to exchange tensors across Spectrum-X leaf switches with high speed and low latency. Which configuration goal should engineers pursue?
- A transit ML cluster enables RoCE on Spectrum-X switches but training still stalls under load. Host PFC and priority settings were never aligned. What is the most likely gap?
- A city GPU fabric wants lower latency for inference backends that currently use TCP sockets across Spectrum-X. Which approach best meets that networking goal?
- After Spectrum-X leaf upgrades, a county clinic telehealth fabric must confirm RoCE still works end to end. What should operators do?
- A 311 model farm needs RDMA over Converged Ethernet between GPU workers on Spectrum-X. Which action aligns with that requirement?
- Water-utility analytics nodes need RoCE on Spectrum-X while keeping management Ethernet separate. Which configuration practice is appropriate?
- A civic open-data GPU pod should enable RoCE only where AI workloads need it on Spectrum-X. What scoping decision is correct?
- Fire/EMS forecasting GPUs need confirmed RoCE prerequisites on Spectrum-X before go-live. Which items must be completed for low-latency municipal AI traffic?
- A municipal AI factory sees PFC storms after enabling lossless RoCE on Spectrum-X. Engineers want congestion signaled without endless pause frames. Which action fits?
- Public-health training jobs starve latency-sensitive inference on the same Spectrum-X fabric. What should operators configure?
- A transit fabric sees hot-spot links clogging all-reduce paths on Spectrum-X. Which advanced feature should engineers enable to dynamically avoid congested paths?
- County assessor jobs need proof that PFC and ECN are actually marking traffic on Spectrum-X. What should the team use?
- A school-district GPU lab enables PFC on Spectrum-X without ECN and then sees head-of-line blocking across leafs. What lesson should operators take?
- Emergency-management inference requires strict priority for RoCE queues on Spectrum-X. What should engineers assign and verify?
- City planning digital-twin traffic needs Spectrum-X adaptive routing plus visible telemetry. Which combination meets that observability goal?
- A parks computer-vision fabric runs load tests on Spectrum-X and needs proof that ECN is marking packets under congestion. What should operators validate?
- A city GPU fabric hosts public-health and permitting tenants on the same Spectrum-X underlay. How should operators isolate those tenant AI workloads?
- Transit and water departments share Spectrum-X leaf switches but must not see each other’s AI VRFs. Which approach provides that isolation?
- A county open-data pod leaks routes between research and production EVPN tenants after a change. What is the most likely cause to investigate first?
- Housing authority and courts inference workloads share a Spectrum-X fabric and need tenant separation. Which multi-tenancy mechanism should designers choose?
- A school-district GPU lab dual-homes servers into Spectrum-X and must keep each host in the correct EVPN tenant. What pattern should engineers apply?
- Municipal cybersecurity SOC AI traffic must stay isolated from 311 model serving on the same Spectrum-X fabric. What should designers establish?
- Public-works and parks departments share a Spectrum fabric. What should engineers verify about VRF behavior by default?
- A civic AI factory onboards a new agency tenant on Spectrum-X. How should operators add that tenant safely?
- Before changing Spectrum-X QoS for a public-health fabric, engineers want to catch issues without touching production. What should they do?
- A transit authority wants to know whether RoCE paths survive a Spectrum-X leaf failure before the change window. What should engineers do?
- County IT wants to prototype a new EVPN tenant design for Spectrum-X without risking live switches. Where should they identify issues first?
- A city GPU fabric upgrade plan needs validation as a digital twin before cutover. Which tool fits that Spectrum environment validation?
- Emergency-management planners add GPU racks and want congestion what-ifs before hardware lands. What should they run?
- Water-utility network engineers need peers to review the intended Spectrum-X topology before deployment. What should they build?
- A school-district change window is shortened after a VLAN mistake is found early. Which tool most likely exposed that Spectrum configuration issue before production?
- Public-health all-reduce slows and Spectrum-X in-band telemetry plus WJH show pause-frame bursts on one leaf. What should operators conclude they are doing?
- A transit fabric sees intermittent RoCE drops; WJH events point to a buffer-threshold problem. What skill are engineers applying?
- County assessor jobs fail mid-epoch; engineers need to tie hotspots to discard events on Spectrum-X. Which approach is appropriate?
- A 311 model farm’s leaf shows discard signatures during peak inference. Which Spectrum service should operators use for real-time event analysis?
- Fire/EMS training hits latency spikes and operators need the congested Spectrum-X path identified. What should they use?
- A municipal AI factory NOC finds that blind SNMP polling misses microbursts on Spectrum-X. What should they enable?
- Parks CV pipeline packet loss is suspected; engineers want fabric proof before blaming the training framework. What should they check first?
- City planning jobs recover after engineers fix a PFC problem that WJH highlighted. What does that outcome demonstrate?
- A public-health fabric needs real-time visibility into RoCE latency and congestion across Spectrum-X leafs. Which tool should operators use?
- Transit ML operators want alerts when Spectrum leaf-to-leaf latency exceeds civic SLOs. Which NetQ capability should they configure?
- County IT validates a Spectrum-X change by comparing congestion before and after. Which tool view should they use?
- A city GPU fabric NOC wants real-time Spectrum flow and congestion visibility instead of ad-hoc CLI scrapes. What should it select?
- Emergency-management inference latency regressions need correlation with fabric measurements. Where should operators look first?
- Water-utility engineers want to detect emerging Spectrum-X congestion before analytics jobs fail. What should they apply?
- School-district lab staff onboard a new Spectrum-X leaf and must confirm monitoring coverage. What should they verify?
- A municipal AI factory needs DPU networking services on BlueField cards that front Spectrum-attached GPU servers. What must operators install?
- Public-health hosts need SuperNIC and DPU offloads on their Spectrum path. What installation is a prerequisite?
- Transit GPU nodes fail DPU bring-up on the Spectrum fabric until a missing software stack is corrected. What is the likely gap?
- County clinic fabric operators want consistent DPU software across BlueField cards. What should they standardize?
- A city GPU fabric must confirm software readiness before enabling advanced BlueField packet services beside Spectrum-X. What should operators check?
- Emergency-management DPUs must support host-path networking offload for civic AI hosts. What unlocks those BlueField capabilities?
- Housing authority cluster documentation covers Spectrum-X leaf configs but omits DPU software. What install steps should the runbook add?
- A public-health inference fabric wants host-side offload for advanced packet processing and congestion control on Spectrum-connected servers. What should operators configure?
- Transit AI servers see RoCE bursts destabilize training on Spectrum-connected hosts. Which SuperNIC capability should they enable?
- County assessor telemetry shows host-side bottlenecks limiting civic AI throughput on Spectrum paths. What should engineers adjust?
- A city GPU fabric must distinguish SuperNIC roles from plain ConnectX NIC baselines. What should architects recognize?
- Emergency-management hosts must confirm SuperNIC offloads before peak wildfire-model season. What should operators verify?
- Water-utility SuperNICs must work with Spectrum-X lossless and ECN policy. How should congestion control be set?
- A school-district lab brings up a new multi-GPU inference rack on Spectrum Ethernet. Which SuperNIC action should staff take?
NVIDIA InfiniBand Networking · 60 questions
- A municipal AI factory brings up a new InfiniBand fabric and needs subnet management high availability. What should operators provision?
- Public-health InfiniBand switches need baseline services before GPU hosts attach. What initial steps should operators complete?
- A transit ML cluster finishes InfiniBand leaf provisioning and must prove subnet manager HA. What should operators test?
- County assessor InfiniBand HA subnet managers must cooperate during provisioning. Which prerequisite should documentation include?
- A city GPU fabric replaces a failed InfiniBand switch and must keep subnet manager HA intact. What should operators do?
- An emergency-management InfiniBand rack is onboarded for civic AI. What should initial configuration include?
- Water-utility engineers finish InfiniBand management IP assignment. What should they verify next for HA?
- A school-district InfiniBand lab needs first-time switch bring-up that matches other civic sites. What should staff apply?
- A housing authority expands InfiniBand by adding switches into an existing HA subnet-manager domain. What must operators preserve?
- Courts batch-inference InfiniBand fabric finishes initial provisioning. What should operators confirm next?
- A 311 model-farm InfiniBand bring-up needs a complete initial checklist. What should that checklist cover?
- A parks computer-vision InfiniBand cluster loses its primary subnet manager during a drill. What should operators validate?
- Fire/EMS InfiniBand fabric work is queued for both management HA and later tenancy. What order should operators follow?
- A civic open-data InfiniBand pod finishes HA provisioning. What should on-call documentation include?
- City planning completes InfiniBand initial configuration with an HA subnet-manager pair. What should operators prove?
- A city GPU fabric must isolate public-health and permitting AI tenants on shared InfiniBand. What should operators configure?
- Transit and water AI jobs share InfiniBand hardware but must not communicate across agencies. What should operators assign?
- After a change window, county clinic InfiniBand nodes cannot join the intended tenant. What should operators diagnose?
- Emergency-management sensitive models run on InfiniBand HCAs that must not see every partition. What should operators do?
- A school-district InfiniBand lab wants each class project isolated on shared IB hardware. What should staff use?
- Housing authority and courts share InfiniBand switches and need tenant isolation. What should operators apply as the control plane?
- 311 and SOC AI tenants accidentally share a default InfiniBand PKey. What should operators do?
- Public-works InfiniBand expansion adds new HCA GUIDs. What PKey action should operators take?
- A water-utility InfiniBand fabric reviews how a tenant should join its partition. What should operators choose carefully?
- City planning must add a new InfiniBand PKey without breaking live tenants. What approach should operators take?
- A Fire/EMS InfiniBand security review checks partition layout. What separation should operators confirm?
- A library computer-vision InfiniBand stack claims tenant isolation. How should operators verify it?
- County assessor InfiniBand auditors ask how agencies map to partitions. What should operators maintain?
- Parks InfiniBand tenant onboarding must limit access for a new civic project. What PKey practice should staff apply?
- A municipal AI factory retires a tenant from shared InfiniBand. What should operators do?
- A public-health InfiniBand fabric sees static paths congest during model training. What should operators configure?
- Transit all-reduce traffic saturates a hot InfiniBand link while inference must stay responsive. What should operators apply?
- County assessor InfiniBand jobs improve after congested flows are redistributed. What fabric feature produced that result?
- An emergency-management InfiniBand fabric must differentiate forecast jobs from batch tenants. What should operators configure?
- A city GPU InfiniBand fabric must start a large multi-rail job. What should operators verify first?
- Water-utility InfiniBand congestion persists after partial tweaks. What combined approach should operators take?
- A school-district InfiniBand lab compares fixed routing with adaptive routing under synthetic load. When does adaptive routing help most?
- Housing-authority InfiniBand path flaps decrease after fabric tuning. What should operators keep using?
- A 311 model-farm InfiniBand SLA requires responsive interactive inference queues. What should operators map that SLA to?
- Fire/EMS InfiniBand operations need clear expectations for when paths change under load. What should the NOC document?
- Parks computer-vision InfiniBand training saturates a spine link. What should operators expect adaptive routing to do?
- A civic open-data InfiniBand pod enables QoS and must keep multi-tenant traffic fair. What tuning goal should operators meet?
- City planning changes InfiniBand subnet-manager HA and must keep traffic classes correct. What should operators re-verify?
- Public-works InfiniBand jobs fight unfairly for congested paths. What should operators select to improve multi-job fairness?
- Courts InfiniBand batch windows must not outrank daytime interactive inference. How should operators set QoS?
- A municipal AI factory expands InfiniBand racks and needs live fabric visibility. What should operators use?
- Public-health InfiniBand bandwidth drops during inference. Where should operators look first in UFM?
- Transit ML operators want warning before InfiniBand jobs fail on a dead link. What should they configure?
- A county clinic updates InfiniBand switch firmware and needs proof of impact. What UFM practice should operators follow?
- An emergency-management NOC wants one primary daily view of InfiniBand link and bandwidth health. What should they select?
- A city GPU fabric sees uneven InfiniBand utilization across rails. What should operators inspect in UFM?
- Water-utility on-call repairs an InfiniBand link and needs confirmation it is healthy again. What should they check?
- School-district staff see InfiniBand job slowdowns and need a fabric correlation habit. What should they learn to do?
- A housing-authority InfiniBand expansion may go live only after fabric health looks clean. What acceptance check should operators require?
- A 311 model-farm InfiniBand fabric needs recurring civic ops visibility. What should operators schedule?
- A Fire/EMS InfiniBand fabric sees a degraded optical link before host jobs fail. Where should operators spot it first?
- Parks computer-vision InfiniBand jobs time out while one spine port collapses in bandwidth. What should operators connect?
- A civic open-data InfiniBand pod wants fabric health in the city observability stack. What should operators do?
- City planning runs an InfiniBand maintenance window and must catch unexpected link flaps live. What should operators watch?
- Courts InfiniBand inventory was updated with new switches and HCAs. What should operators validate in UFM?
Kubernetes Integration · 10 questions
- A public-health Kubernetes cluster needs inference pods to use RDMA/InfiniBand interfaces. What should operators deploy?
- A transit ML platform’s GPU pods must attach InfiniBand/RDMA devices in Kubernetes. What deployment should operators choose?
- County assessor Kubernetes nodes need secondary network attachment for RDMA. What should the Network Operator rollout include?
- A city GPU fabric’s Kubernetes AI namespace cannot schedule RDMA workloads yet. What is the deployment gate?
- An emergency-management Kubernetes cluster will launch fabric-attached training jobs. What should operators sequence first?
- After NVIDIA Network Operator install, public-health admins must prove RDMA/IB is ready. What should they verify?
- A transit platform must prove Network Operator wired InfiniBand/RDMA correctly. What validation should operators run?
- A county clinic must declare the Kubernetes fabric ready after Network Operator install. What health check should come first?
- A school-district lab’s Network Operator verification fails because expected network attachments are missing. What should staff diagnose?
- A city GPU fabric upgrades its Kubernetes cluster and must re-check RDMA enablement. What should ops include?
Troubleshooting Tools · 40 questions
- Spectrum-X leafs in a public-health fabric look starved during triage. Which tool should engineers run first for resource allocation?
- Transit ML congestion aligns with exhausted Spectrum forwarding resources. How should operators use cl-resource-query?
- A county assessor change window grows ACLs and scale on Spectrum-X. What capacity habit should operators apply?
- A city GPU fabric suspects Spectrum-X table exhaustion. Which tool should validate resource headroom?
- An emergency-management NOC runbook starts Spectrum-X allocation triage. Which diagnostic should come early?
- A water-utility Spectrum-X leaf fails new EVPN routes after growth. What should cl-resource-query help show?
- A school-district lab asks which utility diagnoses Spectrum-X resource allocation. What should staff identify?
- A 311 model farm alarms on Spectrum-X capacity, but evidence may disagree. How should operators decide?
- Public-health Spectrum drops need the exact discard reason in real time. What should operators use?
- Transit RoCE microbursts vanish in SNMP averages but still hurt AI jobs. What should operators choose for visibility?
- A county clinic correlates AI job failures with Spectrum leaf events. What should operators map first?
- A city GPU fabric missed short-lived congestion with periodic polls. What should operators enable?
- Emergency-management on-call must isolate PFC-related discards on Spectrum. How should WJH be used?
- Water-utility Spectrum triage needs a fast first look at real-time fabric events. What should come first?
- Parks computer-vision latency spikes align with Spectrum uplink symptoms. What should operators correlate?
- A school-district GPU lab updated Spectrum software and needs troubleshooting readiness. What should operators verify?
- A public-health pod must prove paths meet latency budgets before training. What should operators verify?
- Transit inference regresses after fabric work. How should engineers validate low-latency paths?
- A county assessor imagery pipeline fails its SLO. Interconnect checks point where?
- A city GPU fabric acceptance test must prove end-to-end low latency. What should the design cover?
- An emergency-management cluster changed the fabric and must recheck storage paths. What should operators verify?
- Water-utility nodes pass GPU fabric tests but still miss latency budgets. What distinction matters?
- Housing authority training nodes updated NIC firmware. What should operators do next for interconnects?
- A school-district GPU lab needs repeatable interconnect triage. What should operators establish?
- InfiniBand jobs stall in a municipal AI factory and control-plane health is suspect. What should operators use?
- A public-health InfiniBand fabric looks link-up but jobs still fail. What should operators interpret?
- Transit InfiniBand on-call starts a fabric incident. What should come first in triage?
- A county clinic InfiniBand incident was fixed at the subnet manager. How should closure be confirmed?
- Emergency-management InfiniBand alerts need scoped root-cause domains. What should operators use?
- A city GPU InfiniBand fabric needs recurring civic ops visibility into system health. What should operators schedule?
- Water-utility InfiniBand nodes show unexplained resets. What should operators correlate?
- Courts InfiniBand maintenance sees UFM system health worsen mid-window. What should operators do?
- Public-health InfiniBand latency complaints need a measured write path between GPU nodes. Which command fits?
- Transit validates InfiniBand throughput after a spine upgrade. Which command should confirm bandwidth?
- County assessor staff doubt basic InfiniBand reachability between nodes. What should clear that first?
- A city GPU node shows the wrong InfiniBand state in ibstat after a driver update. What should engineers do first?
- Emergency-management InfiniBand shows fabric-wide anomalies across many nodes. Which diagnostic should operators run?
- A water-utility team must validate InfiniBand inventory against the live fabric map. Which tools should they use?
- A school-district lab reseats a suspect InfiniBand cable and needs proof the path improved. What measurement should they compare?
- A 311 model-farm runbook must escalate InfiniBand connectivity checks in a sensible order. Which sequence is appropriate?
Automation and Configuration · 20 questions
- A municipal AI factory wants Spectrum-X baselines that are not one-off CLI snowflakes. What should operators use?
- Public-health Spectrum-X leafs drift until RoCE and QoS settings disagree. How should engineers restore consistency?
- A transit authority upgrades Spectrum-X and must apply structured interface and EVPN blocks. What is the preferred approach?
- A county clinic fabric has border leafs and AI leafs with different Spectrum-X needs. How should NVUE templates be organized?
- A city GPU fabric wants Spectrum-X changes reviewed before they hit production leafs. Where should NVUE templates live?
- Emergency-management Spectrum-X breaks after a failed manual change. What is the best recovery path?
- Water-utility engineers rendered new Spectrum-X settings from NVUE templates. What should they do before production cutover?
- A school-district lab still manages Spectrum-X with ad-hoc CLI. What modernization step should instructors teach?
- Housing-authority Spectrum onboarding requires switches to match the civic standard. What should the gate enforce?
- A parks computer-vision fabric must update Spectrum-X PFC and ECN settings. How should operators make the change?
- A public-health team must create AI VLANs across many Spectrum-X leafs without manual repetition. What should they write?
- Transit ML enablement needs consistent RoCE configuration on Spectrum-X. Which automation approach fits?
- A county assessor fabric needs both AI VLANs and RoCE prerequisites pushed consistently. What should Ansible deliver?
- A city GPU fabric still uses manual Spectrum-X switch checklists for setup. What should replace them?
- Emergency-management change windows must configure RoCE priorities on Spectrum-X safely. What should run?
- A water-utility network-as-code repo must onboard Spectrum-X VLANs the same way every time. What should authors add?
- A school-district lab is developing Ansible for Spectrum RoCE and VLAN tasks. What safety practice should students use first?
- A 311 model farm’s Ansible Spectrum setup playbooks configure VLANs but rarely prove success. What should they add?
- Fire/EMS Spectrum-X expansion adds new leafs that need the standard AI VLANs. How should operators scale provisioning?
- A civic open-data fabric must run RoCE Ansible playbooks against the right Spectrum roles. What should documentation define?
These questions are original practice material and are NOT actual exam questions or brain-dump content. All vendor marks are trademarks of their respective owners. This site is not affiliated with, endorsed by, or sponsored by the exam vendor.