H100, NVIDIA Data Center GPU with Hopper data center GPU for accelerated AI and high-performance computing

  • 100% Genuine
    100% Genuine
    Original, factory-verified
    3-Year Warranty
    3-Year Warranty
    Extended, enterprise-grade
    23 Years Expertise
    24 Years Expertise
    Proven, deployment-focused
  • Condition
    • Original New
    expert
    Certified Expertise Builds Trust
    CCIE · JNCIE · NSE7 · ACDX · HPE Master ASE · Dell Server/AI Expert
    • Faster Delivery, Globally Ready

      Faster Delivery, Globally Ready

      20+ years of supply chain expertise for faster multi-stop delivery with smooth customs.

      48H Worldwide Shipping
      10+ Global Warehouses
    • Keep Projects On Time

      Keep Projects On Time

      Reliable availability and real-time shipment visibility to keep every critical IT project on schedule.

    • Leading Global Distributor Since 2002

      Leading Global Distributor
      Since 2002

      A global ICT bridge delivering authentic hardware and trusted service for over 20 years.

        21500+ Enterprise Clients Served200+ Countries & Regions Covered
    • Worry-Free Coverage with RSCare

      Worry-Free Coverage with RSCare

      3-Year warranty for enterprise hardware, including advanced RMA, diagnostics, spare parts support, and OEM service compatibility.

      Cisco SmartNet
      HPE Foundation Care
      Dell ProSupport
      Juniper Care
      Fortinet Support
    • Your Trusted Genuine Source

      Your Trusted Genuine Source

      Authorized partnerships with 30+ brands, including Cisco, HPE, Dell, Juniper, and Fortinet.

      100% Authentic Products
      100,000+ Multi-Brand SKUs Available
    Vista general

    NVIDIA Mellanox H100 Vista general

    NVIDIA H100 is a Hopper-generation data center GPU for large-scale AI and high-performance computing. Because the H100 family includes different form factors and memory/power configurations, the PCIe, SXM or NVL variant must be confirmed before ordering.

    Key Specifications

    GPU Architecture

    NVIDIA Hopper with Transformer Engine and second-generation Multi-Instance GPU

    Memory Configuration

    80GB HBM3 for H100 SXM/PCIe; H100 NVL uses a separate 94GB configuration

    Memory Bandwidth

    Up to 3.35 TB/s for SXM, 2 TB/s for PCIe and 3.9 TB/s for NVL

    AI and HPC Performance

    Up to 3,958 FP8 TFLOPS and 3,958 INT8 TOPS with sparsity for H100 SXM

    MIG and Interconnect

    Up to 7 MIGs and NVIDIA NVLink / PCIe Gen 5 connectivity

    Power and Form Factor

    Up to 700W SXM, 350W PCIe or 350–400W NVL; form factor varies by version

    Best Fit Applications

    Large-Scale LLM Training

    Large-Scale LLM Training

    Hopper Tensor Cores and Transformer Engine accelerate large model training and mixed-precision AI workloads.

    Multi-GPU Inference

    Multi-GPU Inference

    NVLink, PCIe Gen 5 and MIG support scalable inference, multi-tenant deployment and GPU partitioning.

    Scientific Computing and HPC

    Scientific Computing and HPC

    High FP64, FP32 and Tensor Core throughput supports scientific simulation, analytics and high-performance computing.

    Cloud and Data Center AI

    Cloud and Data Center AI

    The H100 platform is designed for server and data-center deployments rather than desktop graphics systems.

    Why H100

    Hopper Transformer Engine

    Hopper Transformer Engine

    The Transformer Engine and Tensor Core acceleration support high-throughput mixed-precision training and inference.

    Flexible Data Center Configurations

    Flexible Data Center Configurations

    SXM, PCIe and NVL variants support different server, interconnect, memory and power requirements.

    MIG and NVLink Scale

    MIG and NVLink Scale

    MIG enables partitioned GPU service, while NVLink supports tightly coupled multi-GPU systems.

    Compare Related Models

     
    NVIDIA H100NVIDIA H100
    GPU ArchitectureHopperAmpere
    Memory80GB HBM3 for SXM/PCIe; 94GB for H100 NVL80GB HBM2e
    Memory Bandwidth3.35 TB/s SXM; 2 TB/s PCIe; 3.9 TB/s NVL2,039 GB/s SXM; 1,935 GB/s PCIe
    AI / Compute PerformanceUp to 3,958 FP8 TFLOPS / 3,958 INT8 TOPS with sparsityUp to 624 INT8 TOPS PCIe / 1,248 INT8 TOPS SXM with sparsity
    Max TDPUp to 700W SXM; 350W PCIe; 350–400W NVL400W SXM; 300W PCIe
    Form FactorSXM or PCIe dual-slotSXM or PCIe dual-slot
    InterconnectNVLink and PCIe Gen 5NVLink and PCIe Gen 4
    MIG SupportUp to 7 MIGsUp to 7 MIGs @ 10GB
    Best Fit ScenarioLLM training, multi-GPU inference and scientific computingAI training, inference, data analytics and HPC
    Target CustomerCloud providers, AI companies, research institutions and HPC operatorsOrganizations with established A100 infrastructure, research and enterprise AI teams
    Not Recommended ForDesktop graphics, low-density systems or workloads that do not need data-center accelerationNew deployments requiring Hopper performance or the latest AI features

    Certificaciones de calidad

    • ISO 14001
    • ISO 9001
    • RoHS
    • REACH
    • CE
    • FCC
    Especificación

    NVIDIA Mellanox H100 Especificación

    General
    BrandNVIDIA
    Product CategoryData Center GPU
    ModelH100
    Variant / Form FactorPCIe, SXM or H100 NVL; confirm the exact configuration before ordering
    Architecture and Compute
    ArchitectureNVIDIA Hopper
    FP64 Performance24 TFLOPS PCIe; 30–34 TFLOPS SXM depending on specification source
    FP32 Performance48 TFLOPS PCIe; 60–67 TFLOPS SXM depending on specification source
    FP8 Tensor PerformanceUp to 3,200 TFLOPS PCIe, 4,000 TFLOPS SXM and 3,341 TFLOPS H100 NVL with sparsity
    INT8 Tensor PerformanceUp to 3,200 TOPS PCIe, 4,000 TOPS SXM and 3,341 TOPS H100 NVL with sparsity
    Memory and Interconnect
    GPU Memory80GB HBM3 for H100 SXM/PCIe; 94GB HBM3 for H100 NVL
    GPU Memory Bandwidth3.35 TB/s SXM; 2 TB/s PCIe; 3.9 TB/s NVL
    InterconnectNVIDIA NVLink; PCIe Gen 5
    Multi-Instance GPUUp to 7 MIGs @ 10GB each for H100 SXM/PCIe; up to 7 MIGs @ 12GB each for H100 NVL
    Power and System
    Maximum TDPUp to 700W SXM; 350W PCIe; 350–400W NVL
    Decoder7 × NVDEC and 7 × JPEG decoder engines
    Server OptionsNVIDIA HGX H100, NVIDIA DGX H100 and NVIDIA-Certified Systems
    Still unsure about your configuration? Talk to an engineer for assistance.

    Certificaciones de calidad

    Need Help? Technical Experts Available Now.

    • +1-626-655-0998 (USA)
      UTC 15:00-00:00
    • +852-2592-5389 (HK)
      UTC 00:00-09:00
    • +852-2592-5411 (HK)
      UTC 06:00-15:00
    Need Help? Technical Experts Available Now.
    Garantía
    • Your Trusted Global ICT Partner

      Delivering reliable, cross-border fulfillment with 20+ years of expertise, backed by rigorous QC, fast logistics, and worldwide warehousing.
      Global warehousing
      Global warehousing
      10+ hubs positioned for fast international dispatch and stable availability.
      Seamless cross-border
      Seamless cross-border
      QC-verified packing and predictable delivery via FedEx, DHL, UPS.
      Your Trusted Global ICT Partner
    • ICT Tools Hub for Confident Monitoring

      ICT Tools Hub for Confident Monitoring

      Stay informed on product lifecycle status and real-time inventory availability—supporting confident planning, sourcing, and operational execution.
      EOL/EOSL Checker
      Verify product lifecycle status to plan purchases and avoid EOL/EOSL risks.
      Request VIP Inventory Portal Access
      Connect via live chat for VIP access to real-time inventory on priority SKUs.
    • Certified Experts Across Leading Vendors

      Professional guidance for network, security & data center builds, delivered by certified experts with extensive, hands-on deployment experience.
      Multi-vendor engineering
      Multi-vendor engineering
      Certified expertise across Cisco, HPE, Aruba, Dell, Juniper, Huawei and Fortinet.
      Deployment-ready guidance
      Deployment-ready guidance
      Faster, safer rollouts with project-focused engineering guidance.
      Certified Experts Across Leading Vendors

    The Choice of 21,500 +Global Customers

    • Ford
    • Disney
    • Nestle
    • Toyota
    • Walmart
    • Philips
    • Ups
    • Huatt
    • Hsbc
    • Novartis
    • Siemens Healthineers
    • Bt
    • Reference Johnson
    • T
    • Att
    • Lgu
    • Sk Telecom
    • Veerizon
    Industry-Leading Warranty Coverage You Can Trust
    • 100% Low Price Guarantee
      100% Low Price Guarantee
      Router-switch.com offers genuine, high-quality ICT products at competitive wholesale prices—ensuring exceptional value without compromising quality.
    • 100% Quality Assurance
      100% Quality Assurance
      All items are brand new, original, and factory-sealed.
      Upon request, each unit can be fully tested and verified by a NVIDIA Mellanox-certified engineer to ensure perfect working condition before shipment.
    • 100% Money-Back Guarantee
      100% Money-Back Guarantee
      If you are not satisfied with your purchase, you may request an exchange or full refund. Our customer service team will guide you through the return process quickly and smoothly.
      Visit Política de devolución page for more information.
    • Free NVIDIA Mellanox Apoyo técnico
      Our NVIDIA Mellanox-certified experts provide free professional support via phone, email, online chat, or remote login—helping you deploy and troubleshoot with confidence.
    • Flexible & Secure Payment and Shipping
      Flexible & Secure Payment and Shipping
      We support multiple secure payment and logistics options, backed by industry-leading verification:
      • Pagar por transferencia bancaria.
      • Paga con tu paypal.
      • Paga con tu tarjeta de crédito Visa.
      • Paga con tu tarjeta de crédito Mastercard.
      • Paga con tu tarjeta Discover.
      • Pay by American Express.
      Más
      • mcafee-secure
      • verisign-secured
      • pci-dss
      • ssl-encryption
      Visit How to Pay page for more information
    Warranty Coverage & RMA Instructions
    • Extended, Worry-Free Warranty Coverage
      Extended, Worry-Free Warranty Coverage
      • Factory New Sealed Enterprise Hardware: 3 Years
      • Factory New Sealed Consumer Devices: 1 Year
      Visit Política de garantía page for more information
    • How to Get Support
      How to Get Support
      If you encounter issues with products, please reach out to your dedicated sales representative. They will assist with diagnostics, remote troubleshooting, and RMA processing if necessary.
      Visit Contáctenos page for more support options.

    Best-Fit Solutions for Your Network

    Deploy DeepSeek-V3/R1: Powerful AI, Low Cost

    Deploy DeepSeek-V3/R1: Powerful AI, Low Cost

    Expert-configured AI solution: Same power, 90% savings—multi-scenario flexibility with Dell/Huawei/Lenovo options.

    AI
    Beyond Bandwidth: The 100G+ Data Center Architecture

    Beyond Bandwidth: The 100G+ Data Center Architecture

    The must-have 100G foundation - AI-ready growth, zero-latency performance

    Data Center
    400G/800G Ethernet Switch: Maxmize Margins via AI-Ready Solutions

    400G/800G Ethernet Switch: Maxmize Margins via AI-Ready Solutions

    High-Profit data center switches from Cisco, Huawei, Mellanox & Juniper.

    Ethernet Switch