{"product_id":"nvidia-900-2g133-0300-130-a40-48gb-gddr6-gpu-pci-e-4-0","title":"NVIDIA 900-2G133-0300-130 A40 48GB GDDR6 GPU PCI-E 4.0 P54743-001","description":"\u003cdiv class=\"c4-rich\"\u003e\n\u003cp\u003eThe NVIDIA 900-2G133-0300-130 is an A40 data center GPU with 48 GB of GDDR6 ECC memory on a PCIe 4.0 interface, also listed under HPE number P54743-001. Built on the NVIDIA Ampere architecture, the A40 targets visual computing in the data center: virtual workstations, rendering, and graphics-heavy virtual desktops, with Tensor Core support for AI inference and compute workloads.\u003c\/p\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eAt a glance\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003e48 GB GDDR6 ECC, 696 GB\/s bandwidth\u003c\/li\u003e\n\u003cli\u003ePCIe 4.0 x16, dual slot, passive cooling, 300 W\u003c\/li\u003e\n\u003cli\u003eNVIDIA Ampere with RT and Tensor Cores, vGPU ready\u003c\/li\u003e\n\u003cli\u003eNew, factory sealed\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eOutfitting a rack of GPU servers?\u003c\/h3\u003e\n\u003cp\u003eIf your project calls for more A40s than we have listed, we can source additional quantities. A Core 4 sales engineer can also help with the servers, NVLink bridges, and power cabling to go with them. \u003ca href=\"\/pages\/contact?part_number=900-2G133-0300-130\u0026amp;additional_requirements=Product%3A%20NVIDIA%20900-2G133-0300-130%20A40%2048GB%20GDDR6%20GPU%20PCI-E%204.0%20P54743-001%20%28https%3A%2F%2Fwww.core4solutions.com%2Fproducts%2Fnvidia-900-2g133-0300-130-a40-48gb-gddr6-gpu-pci-e-4-0%29\"\u003e\u003cstrong\u003eTalk to a sales engineer\u003c\/strong\u003e\u003c\/a\u003e or call 1-855-55-CORE4.\u003c\/p\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eTechnical specifications\u003c\/h3\u003e\n\u003ctable style=\"border-collapse:collapse;width:100%;margin:0 0 1em\"\u003e\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eGPU architecture\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eNVIDIA Ampere (2nd-gen RT Cores, 3rd-gen Tensor Cores)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eGPU memory\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e48 GB GDDR6 with ECC\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eMemory bandwidth\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e696 GB\/s\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eHost interface\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003ePCIe 4.0 x16 (64 GB\/s)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eNVLink\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e2-way, 112.5 GB\/s bidirectional; up to 96 GB combined memory\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eDisplay outputs\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e3x DisplayPort 1.4 (disabled by default)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eMax power consumption\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e300 W\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eForm factor\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eDual slot, 4.4 in (H) x 10.5 in (L)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eCooling\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003ePassive\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\" style=\"text-align:left;vertical-align:top;font-weight:600;padding:6px 16px 6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003eVideo engines\u003c\/th\u003e\n\u003ctd style=\"padding:6px 0;border-bottom:1px solid rgba(128,128,128,.2)\"\u003e1x NVENC, 2x NVDEC (AV1 decode)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\u003c\/table\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eCompatibility\u003c\/h3\u003e\n\u003cp\u003eThe A40 is a full-height, dual-slot PCIe 4.0 card with a passive heatsink, so it relies on the server's chassis fans for cooling. Plan for a server that supports a 300 W, double-wide GPU in a PCIe x16 slot, with the airflow and GPU power cabling that the platform calls for.\u003c\/p\u003e\n\u003cp\u003eTwo A40 cards can be linked with a 2-way NVLink bridge to pool memory up to 96 GB. For virtualization, the card supports NVIDIA Virtual PC, Virtual Applications, RTX Virtual Workstation, Virtual Compute Server, and NVIDIA AI Enterprise.\u003c\/p\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eCondition\u003c\/h3\u003e\n\u003cp\u003eThis A40 is new and factory sealed: it ships in the original manufacturer packaging with the factory seal intact, ready to go straight into your build.\u003c\/p\u003e\n\u003cp\u003eIt is covered by Core 4's 1-year advanced replacement warranty. If the card fails within a year, we ship a replacement right away, before the failed unit comes back to us.\u003c\/p\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eLifecycle\u003c\/h3\u003e\n\u003cp\u003eThe A40 is part of NVIDIA's Ampere generation, with second-generation RT Cores and third-generation Tensor Cores.\u003c\/p\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eDatasheets and documentation\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/www.nvidia.com\/en-us\/data-center\/a40\" target=\"_blank\" rel=\"noopener\"\u003eNVIDIA A40 Data Center GPU for Visual Computing\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/images.nvidia.com\/content\/Solutions\/data-center\/a40\/nvidia-a40-datasheet.pdf\" target=\"_blank\" rel=\"noopener\"\u003eNVIDIA A40 datasheet\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/www.nvidia.com\/content\/dam\/en-zz\/Solutions\/Data-Center\/a40\/NVIDIA%20A40%20Product%20Brief.pdf\" target=\"_blank\" rel=\"noopener\"\u003eNVIDIA A40 Product Brief\u003c\/a\u003e\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3 style=\"margin:1.4em 0 0.35em\"\u003eCommon questions\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eWhy don't the A40's DisplayPort outputs show a signal?\u003c\/strong\u003e The A40 ships configured for virtualization with its three DisplayPort 1.4 connectors disabled. NVIDIA's management software tools can switch the physical outputs on when you need direct display output.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eCan I combine two A40 cards for more GPU memory?\u003c\/strong\u003e Yes. Two A40s connected with a 2-way NVLink bridge get 112.5 GB\/s of bidirectional bandwidth between them and can scale to 96 GB of combined memory. The bridge is a separate part.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eDoes the A40 support MIG partitioning?\u003c\/strong\u003e No. The A40 does not support Multi-Instance GPU. It is shared across users with NVIDIA vGPU software such as RTX Virtual Workstation or Virtual PC instead.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eDoes the A40 need a server with GPU airflow?\u003c\/strong\u003e Yes. The card is passively cooled and draws up to 300 W, so it needs a server designed to push enough air through a dual-slot GPU. It is not suited to a desktop case without directed airflow.\u003c\/p\u003e\n\u003c\/div\u003e","brand":"NVIDIA","offers":[{"title":"Default Title","offer_id":68095676219627,"sku":"nvidia 900-2g133-0300-130 new","price":4495.0,"currency_code":"USD","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0843\/8769\/2779\/files\/1_f0868ad2-9f7b-4534-9c3c-f075408a1a20.jpg?v=1793191269","url":"https:\/\/www.core4solutions.com\/products\/nvidia-900-2g133-0300-130-a40-48gb-gddr6-gpu-pci-e-4-0","provider":"Core 4 Solutions","version":"1.0","type":"link"}