← Back to Glossary
What Is NVLink?_
NVLink is NVIDIA's proprietary high-bandwidth interconnect that enables direct GPU-to-GPU communication without routing through the CPU or PCIe bus. NVLink 5.0 (used in GB200/GB300) provides 1.8 TB/s bidirectional bandwidth per GPU. NVLink connections within an NVL72 rack are internal, but external NVLink Switch configurations require precise cable routing.
Technical Details
NVLink has evolved across multiple generations: NVLink 3.0 (A100, 600 GB/s), NVLink 4.0 (H100/H200, 900 GB/s), and NVLink 5.0 (GB200/GB300, 1.8 TB/s). Within an HGX H100 baseboard, NVLink connects 8 GPUs through NVSwitch. In NVL72 rack systems, NVLink 5.0 creates a unified fabric across all 72 GPUs with 130 TB/s total bandwidth. The NVLink connections within the rack are internal (handled by NVSwitch silicon), meaning deployment teams do not route external NVLink cables for NVL72 systems. However, for multi-rack NVLink domains using NVLink Switch, external optical cables must be precisely routed between racks.
How Leviathan Systems Works with NVLink
Leviathan Systems verifies NVLink interconnect functionality as part of our commissioning process, ensuring all GPU-to-GPU links are operational and performing at specification before handoff.
Related service
Network Testing & Commissioning →Related Terms
Appears In
White-Space Planning for Liquid-Cooled GPU HallsWarm-Water Cooling for GPU Clusters: Why Hotter Water WinsTwo-Phase vs Single-Phase Direct Liquid Cooling for GPUsTransceiver Breakout & Splitter Cabling for GPU FabricsThermal Soak & Burn-In During Commissioning: What to WatchNVIDIA Spectrum-X vs Quantum InfiniBand: The Cabling and Optics ViewGetting Ready for Rubin / VR200: Deployment-Readiness Planning NowRigging & Lift Plans for Heavy GPU Racks: Moving Them Without IncidentRFP & SOW Checklist for a GPU Build: Scope It So Nothing Falls ThroughReceiving & Staging GPU Equipment: The Dock-to-Row Logistics PlanRear-Door Heat Exchangers vs Direct Liquid Cooling: Choosing Your Path to High DensityRaised Floor vs Slab for High-Density GPU HallsRack PDU Selection & Metering for GPU Racks: Outlet Count, Phase, TelemetryThe Power-On Walkdown: A Step-by-Step Energization ProcedureScaling from Pilot to Production GPU Cluster: What Breaks and What to PlanOptical Transceiver Handling & Cleanliness on the FloorNCCL All-Reduce Validation as a Cluster Acceptance GateCoordinating a Multi-Site GPU Rollout: Standardize Once, Deploy EverywhereMEP Coordination for AI Data Halls: Where Mechanical, Electrical, and IT CollideBack-End vs Front-End Network Build-Out for GPU ClustersBuilding the Leaf-Spine Cable Plant for a GPU FabricInfiniBand NDR/XDR vs RoCE: What Changes for the Cable PlantHow to Choose a GPU Deployment Partner: A Buyer's ChecklistPre-Terminated Trunks vs Field Termination for GPU FabricHarmonics & Power Quality at GPU Density: Taming the Switch-Mode LoadH200 Deployment Guide: What's Different from H100 on the FloorH100 / HGX H100 Deployment Checklist: Power, Cooling, and Cabling DemandsThe GPU Deployment Site Survey: What to Measure Before You CommitGPU Data Center Deployment in Phoenix, Arizona: Hiring the Build CrewGPU Data Center Deployment in Ohio: Staffing the Physical BuildGPU Data Center Deployment in Northern Virginia: Who Does It and How to HireGPU Data Center Deployment in Georgia: Who Builds It and How to HireGH200 Grace Hopper Deployment Guide: Cabling and Cooling a Superchip NodeGB200 NVL72 Deployment Deep Dive: Liquid Loop, Busbar, and SpineFloor Loading for NVL72 Racks: Will Your Slab Hold ~1.4 Tonnes?Fiber Certification Workflow: From Splice to Signed Acceptance ReportFacility Water Loop Design for AI Halls: TCS to FWS Done RightDe-Racking & Decommissioning a GPU Cluster for MigrationDAC vs AOC vs AEC: Choosing GPU Rack Interconnect CablesContainment & Pathway Build-Out for a GPU RowCondensation & Dew-Point Control in Liquid-Cooled GPU HallsCommissioning Levels L1–L5 for GPU Data Centers, ExplainedB300 Deployment Guide: The Physical-Layer Step Up from B200B200 / HGX B200 Deployment Guide: Power, Thermal, and Rack DemandsThe As-Built & Handoff Package Every GPU Deployment Should DeliverArmored vs Standard Fiber in the GPU Data Hall: When Ruggedization PaysAir-Cooled vs Liquid-Cooled GPU Platforms: Picking the Variant for Your SiteWriting an Acceptance Test Plan (ATP) for a GPU Cluster415V vs 480V Distribution in AI Halls: The High-Density Power Decision400G vs 800G vs 1.6T Optics: Selecting Transceivers for AI FabricSingle-Mode vs Multimode for AI Fabric: The Cable-Plant DecisionCable Pathways & Containment for GPU Rooms: Overhead Tray, Ladder, Fiber RunnerPatch-Panel & Cassette Design for GPU Halls: Breakout, Density, ServiceabilityNVIDIA B200 vs GB200: HGX vs Rack-Scale, and What Changes to DeployMPO Trunk Planning & Fiber-Count Math for GPU HallsHot Aisle vs Cold Aisle Containment at GPU DensityAMD MI300X vs NVIDIA H100: Infrastructure & Deployment DifferencesNVIDIA GB300 NVL72 Explained: Specs, Power, and What It Takes to DeployDell vs Supermicro vs HPE GPU Servers: A Deployment ComparisonData Center Migration for AI Infrastructure: A Practical Field GuideNVIDIA H100 vs H200 vs B200: What Changes for DeploymentGoogle TPU vs NVIDIA GPU: What It Means for Your AI Infrastructure BuildDirect-to-Chip vs Immersion Liquid Cooling for GPU Data CentersData Center Structured Cabling Standards: TIA-942, TIA-606-C, BICSIData Center Rack-and-Stack Services for GPU Builds: What's IncludedCommon GPU Deployment Mistakes — and How to Avoid ThemWho Deploys GB200 / GB300 NVL72 Infrastructure?How Long Does GPU Cluster Deployment Take?GPU Data Center Deployment in Texas: Who Does It and How to HireGPU Commissioning & Acceptance: What to Demand Before You Sign OffSite Readiness Before the GPUs Arrive: Power, Cooling, Floor, PathwaysIn-House vs. Outsourced GPU Deployment: How to DecideGPU Rack Assembly: What Drives the CostSpectrum-X vs InfiniBand: What's Different for the Cable PlantHGX vs DGX: What's Different When You Deploy ThemPre-Power Inspection: The Walkdown Before Energizing a GPU HallNVLink Spine Cartridge & Copper Backplane Handling: Field ProcedureGB300 NVL72 Deployment: Power, Cooling, and the Cable PlantGB200 vs GB300 NVL72: What Changes for DeploymentNCCL Bandwidth Validation: Proving a GPU Fabric Before ProductionGPU Rack Receiving, Staging & Lift Plan: Moving ~1,360 kg Racks Without DamageAir-to-Liquid Cooling Retrofit: The Install SideThermal Burn-In for GPU Clusters: Duration, Watch Items, Pass/FailStructured Cabling QA/QC for GPU Racks: Bend Radius, Slack, Torque, DressingOTDR & Insertion/Return-Loss Testing for GPU Cluster FiberMPO Polarity (Method A/B/C) for GPU Fabric — and the #1 Cause of Dead Links