Career upgrade: Learn practical AI skills for better jobs and higher pay.
Level up
All Practice Exams

100+ Free 300-640 DCAI Practice Questions

Pass your Implementing Cisco Data Center AI Infrastructure (DCAI 300-640) exam on the first try — instant access, no signup required.

✓ No registration✓ No credit card✓ No hidden fees✓ Start practicing immediately
100+ Questions
100% Free
1 / 100
Question 1
Score: 0/0

Which Cisco AI pod architecture component is responsible for storage traffic isolation in a GPU cluster?

A
B
C
D
to track
2026 Statistics

Key Facts: 300-640 DCAI Exam

55-65

Exam Questions

Cisco

90 min

Exam Duration

Cisco

$300

Exam Fee

Pearson VUE

3 years

Certification Validity

Cisco

30%

Largest Domain (tie)

Architecture + Deployment

Feb 2026

Exam Available Since

Cisco Live Amsterdam

The 300-640 DCAI exam has 55-65 questions in 90 minutes with a scaled passing score. Key domains: AI Infrastructure Components & Architecture (30%), AI Infrastructure Deployment & Data Management (30%), AI Fundamentals (20%), and Operations & Troubleshooting (20%). Exam fee is $300. Valid 3 years. Available at Pearson VUE test centers and online proctored.

Sample 300-640 DCAI Practice Questions

Try these sample questions to test your 300-640 DCAI exam readiness. Each question includes a detailed explanation. Start the interactive quiz above for the full 100+ question experience with AI tutoring.

1Which transport protocol is required for RoCEv2 to operate correctly over a lossless Ethernet fabric?
A.TCP
B.UDP
C.GRE
D.MPLS
Explanation: RoCEv2 encapsulates RDMA over UDP/IP, enabling it to traverse routed networks while maintaining low-latency characteristics.
2In a BGP-EVPN VXLAN fabric, which BGP address family is used to carry MAC/IP advertisement routes?
A.IPv4 unicast
B.L2VPN EVPN
C.VPNv4
D.IPv6 unicast
Explanation: The L2VPN EVPN address family carries EVPN route types including MAC/IP advertisement (Type 2) routes across the BGP fabric.
3What is the primary function of the Nexus 9000 series leaf switch in an AI pod Clos fabric?
A.Act as BGP route reflector
B.Provide ToR connectivity for compute nodes
C.Terminate VXLAN tunnels at the WAN edge
D.Run the DCNM controller software
Explanation: In a Clos fabric AI pod design, Nexus 9000 leaf switches provide Top-of-Rack (ToR) connectivity for GPU servers and compute nodes.
4Which congestion control mechanism is commonly deployed in RoCEv2 AI fabrics to prevent buffer overflow without relying solely on PFC?
A.DCQCN
B.RED
C.WRED
D.FIFO queuing
Explanation: DCQCN (Data Center Quantized Congestion Notification) combines ECN and rate-based control to manage congestion in RoCEv2 fabrics, reducing reliance on PFC.
5In an AI GPU cluster, which network tier is responsible for the high-bandwidth collective communication between GPUs?
A.Management network
B.Out-of-band network
C.High-performance compute fabric
D.Storage replication network
Explanation: The high-performance compute fabric (often InfiniBand or RoCEv2 Ethernet) handles all-reduce and other collective operations between GPUs during training.
6What Nexus 9000 feature allows a single physical port to be divided into multiple logical subinterfaces for VXLAN routing?
A.vPC
B.FEX
C.SVI
D.Sub-interface
Explanation: Sub-interfaces on Nexus 9000 allow a single physical interface to carry multiple VLANs as routed interfaces, supporting VXLAN inter-VLAN routing use cases.
7Which InfiniBand feature provides hardware-level multicast for GPU all-reduce operations?
A.RDMA write
B.Subnet manager
C.Multicast groups
D.Queue pairs
Explanation: InfiniBand multicast groups allow the subnet manager to configure hardware multicast trees, enabling efficient all-reduce collectives across GPU nodes.
8What is the recommended MTU size for a RoCEv2 lossless Ethernet AI fabric to accommodate RDMA frames?
A.1500 bytes
B.4096 bytes
C.9000 bytes jumbo
D.64000 bytes
Explanation: Jumbo frames (9000 bytes MTU) are required in RoCEv2 fabrics to reduce per-packet overhead for large RDMA transfers between GPUs.
9In a BGP-EVPN spine-leaf fabric, which EVPN route type is used to signal IP prefix reachability for inter-subnet routing?
A.Type 1
B.Type 2
C.Type 3
D.Type 5
Explanation: EVPN Route Type 5 (IP Prefix Route) is used to advertise IP prefixes for inter-subnet routing in BGP-EVPN fabrics.
10Which Cisco AI pod architecture component is responsible for storage traffic isolation in a GPU cluster?
A.Spine switch
B.Leaf ToR switch
C.Storage leaf
D.Out-of-band management switch
Explanation: A dedicated storage leaf switch is used in AI pod designs to isolate high-throughput storage traffic (e.g., NFS/NVMe-oF) from compute fabric traffic.

About the 300-640 DCAI Exam

The 300-640 DCAI exam validates advanced skills in designing, implementing, and troubleshooting Cisco data center AI infrastructure. It became available for testing in February 2026. The exam covers AI/ML workload fundamentals, GPU cluster networking (RoCE/RoCEv2), Cisco AI Pod architecture, UCS compute configuration, high-performance fabric design with congestion control, Nexus Dashboard orchestration, and operational monitoring. It serves as a concentration exam toward the CCNP Data Center certification.

Questions

60 scored questions

Time Limit

90 minutes

Passing Score

~750-850/1000 (scaled)

Exam Fee

$300 (Pearson VUE)

300-640 DCAI Exam Content Outline

20%

AI Fundamentals and Applications

AI/ML workload types (RAG pipelines, model training, inference serving, generative AI), the AI project lifecycle, infrastructure deployment models (public cloud, hybrid, on-premises, edge), and Cisco AI portfolio components including AI PODs, AI Canvas, and Hyperfabric

30%

AI Infrastructure Components and Architecture

Evaluate network topology options for AI clusters, assess GPU compute deployment requirements, determine storage architecture for AI workloads (NVMe-oF, high-throughput object storage), address power density and sustainability, and design hybrid cloud integration points

30%

AI Infrastructure Deployment and Data Management

Configure lossless Ethernet fabrics using Priority Flow Control (PFC), Explicit Congestion Notification (ECN), and Enhanced Transmission Selection (ETS); deploy RoCE and RoCEv2 for GPU-to-GPU RDMA; configure Cisco UCS for AI compute; deploy fabric automation with Nexus Dashboard, APIC, Hyperfabric, and Intersight

20%

AI Infrastructure Operations and Troubleshooting

Benchmark AI workload performance, monitor infrastructure health using Cisco management tools, track GPU utilization and fabric telemetry, and troubleshoot congestion and connectivity issues in AI clusters

How to Pass the 300-640 DCAI Exam

What You Need to Know

  • Passing score: ~750-850/1000 (scaled)
  • Exam length: 60 questions
  • Time limit: 90 minutes
  • Exam fee: $300

Keys to Passing

  • Complete 500+ practice questions
  • Score 80%+ consistently before scheduling
  • Focus on highest-weighted sections
  • Use our AI tutor for tough concepts

300-640 DCAI Study Tips from Top Performers

1Focus equally on the two 30% domains: AI Infrastructure Components/Architecture and Deployment/Data Management — together they make up 60% of the exam
2Master RoCE/RoCEv2 configuration: understand why lossless Ethernet is required and how PFC, ECN, and ETS work together
3Know Cisco's AI portfolio by name: AI PODs, AI Canvas, Hyperfabric, Nexus Dashboard, Intersight — understand what each tool manages
4Understand GPU cluster network topologies: spine-leaf, rail-optimized, and how InfiniBand compares to Ethernet for AI workloads
5Practice Cisco UCS CLI and Intersight workflows for compute provisioning in AI environments
6Study congestion management in detail — PFC pause frames, ECN marking thresholds, and ETS bandwidth guarantees are high-frequency exam topics
7Use Cisco's free AI infrastructure whitepapers and design guides on cisco.com to understand validated AI Pod architectures
8Allocate 120-160 hours of study; this is an advanced exam requiring hands-on lab or simulator practice

Frequently Asked Questions

What is the Cisco 300-640 DCAI exam?

The 300-640 DCAI (Implementing Cisco Data Center AI Infrastructure) is a 90-minute concentration exam for CCNP Data Center certification. It validates expertise in deploying and managing Cisco AI infrastructure including GPU clusters, RoCE networking, UCS compute, and AI Pod architecture. The exam became available in February 2026.

How much does the 300-640 DCAI exam cost?

The 300-640 DCAI exam costs $300 USD. It is administered by Pearson VUE at test centers worldwide and through online proctored testing. Exam vouchers may be available through Cisco authorized learning partners.

What are the prerequisites for the 300-640 DCAI exam?

There are no hard prerequisites to sit the 300-640 DCAI exam. However, to earn the CCNP Data Center certification you must also pass the 350-601 DCCOR core exam. Cisco recommends 3-5 years of data center experience and completion of the DCAI authorized training course before attempting the exam.

What topics are covered on the 300-640 DCAI exam?

The 300-640 DCAI exam covers four domains: AI Infrastructure Components and Architecture (30%), AI Infrastructure Deployment and Data Management (30%), AI Fundamentals and Applications (20%), and AI Infrastructure Operations and Troubleshooting (20%). Key technologies include RoCE/RoCEv2, PFC, ECN, Cisco UCS, Nexus Dashboard, APIC, and Hyperfabric.

How long is the 300-640 DCAI certification valid?

The CCNP Data Center certification (which includes the 300-640 DCAI concentration) is valid for 3 years. You can renew by passing any CCNP-level or higher exam, earning Continuing Education credits, or by retaking the core exam before expiration.

What is the passing score for the 300-640 DCAI exam?

Cisco uses a scaled scoring model and does not publish a fixed passing percentage. Based on available data, the passing score is approximately 750-850 out of 1000. Scores are reported immediately upon completion at Pearson VUE testing centers.