{"product_id":"nvidia-hgx-h100-sxm5-8gpu-board-935242870301000-board-new","title":"NVIDIA HGX H100 SXM5 8‑GPU Board (935‑24287‑0301‑000) Board - New","description":"\u003ch1\u003e\u003cstrong\u003eNVIDIA HGX H100 SXM5 8‑GPU Board (935‑24287‑0301‑000)\u003c\/strong\u003e\u003c\/h1\u003e\n\u003ch2\u003e\u003cstrong\u003eEnterprise-Grade 8‑Way H100 SXM5 AI Accelerator Platform\u003c\/strong\u003e\u003c\/h2\u003e\n\u003chr\u003e\n\u003cp\u003eThe \u003cstrong\u003eNVIDIA HGX H100 SXM5 8‑GPU Board\u003c\/strong\u003e (part number \u003cstrong\u003e935‑24287‑0301‑000\u003c\/strong\u003e) integrates eight \u003cstrong\u003eNVIDIA Hopper‑based H100 SXM5 GPUs\u003c\/strong\u003e, each with \u003cstrong\u003e80 GB HBM3 memory\u003c\/strong\u003e, into a single high-density accelerator board. Built for \u003cstrong\u003eAI training\u003c\/strong\u003e, \u003cstrong\u003einference\u003c\/strong\u003e, and exascale HPC, it supports direct liquid cooling. With \u003cstrong\u003eNVLink\u003c\/strong\u003e at \u003cstrong\u003e900 GB\/s\u003c\/strong\u003e interconnect and 3.35 TB\/s memory bandwidth, this platform delivers unmatched scalability and performance.\u003c\/p\u003e\n\u003chr\u003e\n\u003ch3\u003e\u003cstrong\u003e🔧 Product Specifications for \u003ca href=\"https:\/\/docs.nvidia.com\/dgx\/dgxh100-user-guide\/introduction-to-dgxh100.html\" title=\"NVIDIA HGX H100 SXM5 8‑GPU Board (935‑24287‑0301‑000) - Data Sheet\" rel=\"noopener\" target=\"_blank\"\u003eNVIDIA HGX H100 SXM5 8‑GPU Board (935‑24287‑0301‑000)\u003c\/a\u003e\u003c\/strong\u003e\u003c\/h3\u003e\n\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003e\u003cstrong\u003eFeature\u003c\/strong\u003e\u003c\/th\u003e\n\u003cth\u003e\u003cstrong\u003eDetails\u003c\/strong\u003e\u003c\/th\u003e\n\u003c\/tr\u003e\n\u003c\/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003ePart Name \/ Number\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eNVIDIA HGX H100 SXM5 8‑GPU Board – 935‑24287‑0301‑000\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eForm Factor\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eSXM5 (8 GPUs on single board, direct liquid-cooled)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eGPUs\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003e8 × NVIDIA H100 SXM5, each 80 GB HBM3\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eTotal GPU Memory\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003e640 GB HBM3\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eMemory Bandwidth\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003e3.35 TB\/s per GPU via HBM3, \u003cem\u003eNVLink at 900 GB\/s per NVLink\u003c\/em\u003e\n\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eCUDA Cores (Total)\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003e8 × 16,896 = 135,168 cores\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eTensor Cores (Total)\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003e8 × 528 = 4,224 cores\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eGPU Interconnect\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eNVLink SXM5 — 900 GB\/s GPU ↔ GPU\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eForm Thermal Design Power\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eConfigurable up to 700 W per GPU\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eCooling\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eDirect liquid cooling (equipment not included)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003ePower Interface\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eBoard-level power connectors (server-integrated power delivery)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003ePCIe Host Interface\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003ePCIe Gen 5 x16 interface (board to host)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eMulti‑Instance GPU (MIG)\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eEach H100 supports up to 7 MIG instances\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eUse Cases\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eDistributed AI training, supercomputing, HPC clusters\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003ctd\u003e\u003cstrong\u003eCertifications\u003c\/strong\u003e\u003c\/td\u003e\n\u003ctd\u003eHopper microarchitecture, NVLink-enabled, NVIDIA HGX\/DGX compliant\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003chr\u003e\n\u003ch3\u003e\u003cstrong\u003e❓ Frequently Asked Questions (FAQs)\u003c\/strong\u003e\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eQ1: What is the total GPU memory on this board?\u003c\/strong\u003e\u003cbr\u003eA: The board provides \u003cstrong\u003e640 GB total HBM3 memory\u003c\/strong\u003e—80 GB across 8 SXM5 GPUs.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eQ2: How do the GPUs communicate within the board?\u003c\/strong\u003e\u003cbr\u003eA: Via \u003cstrong\u003eSXM5 NVLink\u003c\/strong\u003e, offering up to \u003cstrong\u003e900 GB\/s per connection\u003c\/strong\u003e, enabling high-speed GPU-to-GPU communication.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eQ3: What cooling solution does this board require?\u003c\/strong\u003e\u003cbr\u003eA: It requires \u003cstrong\u003edirect liquid cooling\u003c\/strong\u003e, typically implemented in purpose-built HX\/L48 chassis. (Board is liquid-cooled ready.)\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eQ4: Is this board plug-and-play with standard servers?\u003c\/strong\u003e\u003cbr\u003eA: No. It's designed for \u003cstrong\u003eHGX-compatible host systems\u003c\/strong\u003e with integrated power, liquid cooling, and server infrastructure.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eQ5: Can I run multi-instance workloads?\u003c\/strong\u003e\u003cbr\u003eA: Yes—each H100 supports up to \u003cstrong\u003e7 MIG instances\u003c\/strong\u003e, suitable for concurrent multi-tenant or mixed workloads.\u003c\/p\u003e","brand":"NVIDIA","offers":[{"title":"Default Title","offer_id":46731775049889,"sku":"NVIDIA-HGX-H100-SXM5","price":2685000.0,"currency_code":"BRL","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/2216\/8673\/files\/NVIDIA-HGX-H100-SXM5-8-GPU-Board-935-24287-0301-000-Board-New.jpg?v=1784705538","url":"https:\/\/www.foxti.com.br\/products\/nvidia-hgx-h100-sxm5-8gpu-board-935242870301000-board-new","provider":"FoxTI","version":"1.0","type":"link"}