{"product_id":"nvidia-h200-nvl-141gb-hbm3e-pcie-50-gpu-900-21010-0040-000","title":"Nvidia H200 NVL 141GB HBM3e PCIe 5.0 GPU - 900-21010-0040-000","description":"\u003ch2\u003eNVIDIA H200 NVL 141GB HBM3e PCIe 5.0 x16 GPU\u003c\/h2\u003e\u003cp\u003eThe NVIDIA H200 NVL (900-21010-0040-000) is NVIDIA's flagship AI and HPC accelerator, featuring 141GB of HBM3e memory — the largest GPU memory capacity available. Built on the Hopper architecture with PCIe 5.0 x16 connectivity, it delivers 4.8TB\/s memory bandwidth and exceptional FP8\/FP16\/BF16 throughput for large language model (LLM) inference, AI training, and high-performance computing workloads.\u003c\/p\u003e\u003ch3\u003eKey Features\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003e141GB HBM3e — highest GPU memory capacity for large model inference\u003c\/li\u003e\n\u003cli\u003e4.8TB\/s memory bandwidth — 76% more than H100 SXM\u003c\/li\u003e\n\u003cli\u003ePCIe 5.0 x16 interface for broad server compatibility\u003c\/li\u003e\n\u003cli\u003eHopper architecture — FP8, FP16, BF16, TF32, FP64 precision support\u003c\/li\u003e\n\u003cli\u003eNVLink bridge support for multi-GPU configurations\u003c\/li\u003e\n\u003cli\u003eDesigned for LLM inference, AI training, and HPC workloads\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003eProduct Specifications\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003e\n\u003cstrong\u003ePart Number:\u003c\/strong\u003e 900-21010-0040-000\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eManufacturer:\u003c\/strong\u003e NVIDIA\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eOEM Status:\u003c\/strong\u003e Retail\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eModel:\u003c\/strong\u003e H200 NVL\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eArchitecture:\u003c\/strong\u003e Hopper (GH100)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eMemory:\u003c\/strong\u003e 141GB HBM3e\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eInterface:\u003c\/strong\u003e PCI-Express 5.0 x16\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eForm Factor:\u003c\/strong\u003e SXM \/ PCIe (verify form factor for this SKU)\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003eProcessor Specifications\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003e\n\u003cstrong\u003eArchitecture:\u003c\/strong\u003e Hopper (GH100)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eCUDA Cores:\u003c\/strong\u003e 16,896\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eTensor Cores:\u003c\/strong\u003e 528 (4th Gen)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eNVLink:\u003c\/strong\u003e 4th Gen NVLink\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eTDP:\u003c\/strong\u003e [Verify from NVIDIA datasheet]\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003ePerformance Specifications\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003e\n\u003cstrong\u003eMemory Bandwidth:\u003c\/strong\u003e 4.8 TB\/s\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eFP8 Tensor:\u003c\/strong\u003e Up to 3,958 TFLOPS\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eFP16 \/ BF16 Tensor:\u003c\/strong\u003e Up to 1,979 TFLOPS\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eFP64 Tensor:\u003c\/strong\u003e Up to 67 TFLOPS\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eInterface Bandwidth:\u003c\/strong\u003e PCIe 5.0 x16 (128 GB\/s bidirectional)\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003eUse Cases\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003eLarge language model (LLM) inference — fits larger models in single GPU\u003c\/li\u003e\n\u003cli\u003eGenerative AI training and fine-tuning\u003c\/li\u003e\n\u003cli\u003eHigh-performance computing (HPC) and scientific simulation\u003c\/li\u003e\n\u003cli\u003eMulti-GPU AI clusters via NVLink\u003c\/li\u003e\n\u003cli\u003eData analytics and recommendation systems\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003eCompatibility\u003c\/h3\u003e\u003cul\u003e\n\u003cli\u003ePCIe 5.0 x16 servers and workstations\u003c\/li\u003e\n\u003cli\u003eCompatible with PCIe 4.0 x16 slots (at reduced bandwidth)\u003c\/li\u003e\n\u003cli\u003eNVIDIA NVLink bridge for multi-GPU configurations\u003c\/li\u003e\n\u003cli\u003eRequires appropriate power delivery — verify server PSU capacity\u003c\/li\u003e\n\u003c\/ul\u003e\u003ch3\u003eFrequently Asked Questions\u003c\/h3\u003e\u003cdl\u003e\n\u003cdt\u003eWhat is the difference between H200 NVL and H100?\u003c\/dt\u003e\n\u003cdd\u003eThe H200 NVL offers 141GB HBM3e vs H100's 80GB HBM2e\/HBM3, with 4.8TB\/s memory bandwidth vs 3.35TB\/s. The larger memory capacity enables inference of significantly larger AI models without model sharding.\u003c\/dd\u003e\n\u003cdt\u003eIs this an SXM or PCIe form factor?\u003c\/dt\u003e\n\u003cdd\u003eThe H200 NVL is available in both SXM and PCIe form factors. Verify the specific form factor for part number 900-21010-0040-000 from NVIDIA's official product page before ordering.\u003c\/dd\u003e\n\u003cdt\u003eDoes this require special cooling?\u003c\/dt\u003e\n\u003cdd\u003eYes, the H200 NVL requires active cooling. It is designed for data center servers with appropriate airflow — not suitable for standard desktop workstations.\u003c\/dd\u003e\n\u003c\/dl\u003e\u003cp\u003eThe NVIDIA H200 NVL 900-21010-0040-000 delivers 141GB HBM3e with 4.8TB\/s memory bandwidth on the Hopper architecture — the definitive GPU for large-scale AI inference, LLM training, and HPC workloads requiring maximum GPU memory capacity.\u003c\/p\u003e","brand":"Nvidia","offers":[{"title":"Default Title","offer_id":48094856020129,"sku":"900-21010-0040-000","price":1979967.0,"currency_code":"BRL","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/2216\/8673\/files\/Nvidia-H200-NVL-141GB-HBM3e-PCIe-5_0-GPU-900-21010-0040-000.jpg?v=1792133554","url":"https:\/\/www.foxti.com.br\/products\/nvidia-h200-nvl-141gb-hbm3e-pcie-50-gpu-900-21010-0040-000","provider":"FoxTI","version":"1.0","type":"link"}