{"product_id":"nvidia-nvlink-bridge-2-way-for-h200-nvl","title":"NVIDIA NVLink Bridge 2-way for H200 NVL","description":"\u003csection class=\"phd-intro\"\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003eData centre GPU accessory \/ NVIDIA NVLink\u003c\/p\u003e\n\u003ch2\u003eLink two H200 NVL cards over NVLink.\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cp\u003eThe NVIDIA 2-way NVLink bridge joins two adjacent H200 NVL PCIe cards so they exchange data over NVLink instead of PCIe. NVIDIA rates the link at up to 900GB\/s per GPU, against 128GB\/s for PCIe Gen5 x16 (both bidirectional). One bridge serves one pair of cards.\u003c\/p\u003e\n\u003cp\u003eThis is NVIDIA part 900-23945-0000-000. It is made for the H200 NVL; H100 NVL, H100 PCIe and A100 PCIe cards use a different bridge.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/section\u003e\n\u003cnav class=\"phd-jump-links\"\u003e\u003ca href=\"#hardware-components\"\u003eProduct details\u003c\/a\u003e \u003ca href=\"#hardware-specification\"\u003eTechnical specification\u003c\/a\u003e \u003ca href=\"#hardware-included\"\u003eIncluded and required\u003c\/a\u003e \u003ca href=\"#hardware-questions\"\u003eBefore you order\u003c\/a\u003e\u003c\/nav\u003e\n\u003cdiv class=\"phd-metrics\"\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e900GB\/s\u003c\/strong\u003e\u003cspan\u003eNVLink per GPU, bidirectional\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e2 cards\u003c\/strong\u003e\u003cspan\u003eAdjacent H200 NVL GPUs\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e1 per pair\u003c\/strong\u003e\u003cspan\u003eOne bridge links each pair\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e49g\u003c\/strong\u003e\u003cspan\u003ePassive bridge\u003c\/span\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003csection class=\"phd-components\" id=\"hardware-components\"\u003e\n\u003ch2\u003eOne wide bridge for each pair of GPUs.\u003c\/h2\u003e\n\u003cdiv class=\"phd-features\"\u003e\n\u003carticle\u003e\n\u003ch3\u003eA single bridge per pair\u003c\/h3\u003e\n\u003cp\u003eThe H100 NVL needs three separate bridges per pair. The H200 NVL instead uses one wide bridge across the top edges of two cards installed side by side. In an eight-GPU server, four 2-way bridges create four NVLink pairs. For a group of four GPUs on NVLink, NVIDIA makes a separate 4-way bridge.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003carticle\u003e\n\u003ch3\u003eGPU-to-GPU traffic off PCIe\u003c\/h3\u003e\n\u003cp\u003eWhen one model or job is split across both GPUs, for example tensor-parallel LLM inference or multi-GPU training, the cards exchange data frequently. The bridge carries that traffic over NVLink, which NVIDIA rates at up to 900GB\/s per GPU (bidirectional). Each card keeps its own 141GB of HBM3e; software that splits the work across both cards can use the faster link between them.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003carticle\u003e\n\u003ch3\u003eFor servers qualified for the H200 NVL\u003c\/h3\u003e\n\u003cp\u003eServer vendors qualify their own versions of this 2-way bridge for particular systems: Lenovo 4X67A97320 in the ThinkSystem SR675 V3 and SR650 V4\/SR650a V4, HPE S4A90C in the ProLiant DL380a Gen12 and DL385 Gen11, and Cisco UCSX-NVL2-H200 in the UCS X580p PCIe Node. Cisco's version is the NVIDIA part in this listing, 900-23945-0000-000. NVIDIA points to its MGX H200 NVL partner and NVIDIA-Certified Systems. The two cards must sit in adjacent slots.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003c\/div\u003e\n\u003c\/section\u003e\n\u003caside class=\"phd-note\"\u003e\n\u003ch3\u003eCheck your server supports the H200 NVL with this bridge.\u003c\/h3\u003e\n\u003cp\u003eThe bridge fits across two adjacent H200 NVL cards and needs at least 2.5mm of clearance above their top edge and at least 2.67mm behind the rear card's PCB. NVIDIA recommends pairing cards attached to the same CPU or PCIe switch. Lenovo and HPE qualify their own part numbers for their servers, so check with your server vendor before fitting this NVIDIA-branded bridge.\u003c\/p\u003e\n\u003c\/aside\u003e\n\u003csection class=\"phd-specs\" id=\"hardware-specification\"\u003e\n\u003cdiv class=\"phd-section-heading\"\u003e\n\u003ch2\u003eTechnical specification\u003c\/h2\u003e\n\u003cp\u003eSpecifications for one bridge\u003c\/p\u003e\n\u003c\/div\u003e\n\u003ctable\u003e\n\u003ccaption\u003e01 \/ Bridge\u003c\/caption\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eProduct\u003c\/th\u003e\n\u003ctd\u003eNVIDIA NVLink bridge, 2-way, passive\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNVIDIA part number\u003c\/th\u003e\n\u003ctd\u003e900-23945-0000-000\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCompatible GPU\u003c\/th\u003e\n\u003ctd\u003eNVIDIA H200 NVL (PCIe, 141GB HBM3e)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCards linked\u003c\/th\u003e\n\u003ctd\u003eTwo, installed side by side in adjacent slots\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eBridges per pair\u003c\/th\u003e\n\u003ctd\u003eOne\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNVLink bandwidth\u003c\/th\u003e\n\u003ctd\u003eUp to 900GB\/s per GPU, bidirectional (NVIDIA rating)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eWeight\u003c\/th\u003e\n\u003ctd\u003e49g\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003ctable\u003e\n\u003ccaption\u003e02 \/ Installation\u003c\/caption\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003ePosition\u003c\/th\u003e\n\u003ctd\u003eAcross the top edges of two adjacent H200 NVL cards\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eClearance\u003c\/th\u003e\n\u003ctd\u003eAt least 2.5mm above the cards' top edge, and at least 2.67mm behind the rear card's PCB\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eRecommended topology\u003c\/th\u003e\n\u003ctd\u003eBoth cards attached to the same CPU or PCIe switch\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eFitting\u003c\/th\u003e\n\u003ctd\u003eClips onto the cards' NVLink connectors after their covers are removed\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eRemoval\u003c\/th\u003e\n\u003ctd\u003eTwo NVIDIA bridge removal tools, releasing both ends together\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003cp class=\"phd-footnote\"\u003eNVIDIA gives up to 900GB\/s for NVLink and 128GB\/s for PCIe Gen5 x16; both are bidirectional totals per GPU. Server vendors also sell the 2-way bridge under their own part numbers: Lenovo 4X67A97320 (NVIDIA 900-23945-2700-000) and HPE S4A90C. Cisco's UCSX-NVL2-H200 is NVIDIA 900-23945-0000-000, the part in this listing.\u003c\/p\u003e\n\u003c\/section\u003e\n\u003csection class=\"phd-supply\" id=\"hardware-included\"\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003eIncluded\u003c\/p\u003e\n\u003ch2\u003eOne 2-way NVLink bridge.\u003c\/h2\u003e\n\u003cp\u003eOne NVIDIA 2-way NVLink bridge for H200 NVL, part number 900-23945-0000-000. Bridge removal tools are not part of this listing.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003ePlan your installation\u003c\/p\u003e\n\u003ch2\u003eWhat your server needs.\u003c\/h2\u003e\n\u003cul\u003e\n\u003cli\u003eTwo NVIDIA H200 NVL PCIe cards for each bridge.\u003c\/li\u003e\n\u003cli\u003eA server that supports the H200 NVL with a 2-way NVLink bridge, with the two cards in adjacent slots; check which bridge part your server vendor supports.\u003c\/li\u003e\n\u003cli\u003eAn NVIDIA data centre driver that supports the H200 NVL.\u003c\/li\u003e\n\u003cli\u003eTwo NVIDIA bridge removal tools if you later need to remove the bridge.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/section\u003e\n\u003csection class=\"phd-faq\" id=\"hardware-questions\"\u003e\n\u003cdetails\u003e\n\u003csummary\u003eHow many bridges do I need?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eOne for each pair of H200 NVL cards: one bridge for two cards, two for four cards set up as two pairs, and four for eight cards. To join four cards in a single NVLink group, use NVIDIA's separate 4-way bridge instead.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eDoes it work with H100 NVL, H100 PCIe or A100 cards?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eNo. This bridge is for the H200 NVL. The H100 NVL, H100 PCIe and A100 PCIe use a different NVLink bridge, NVIDIA part 900-53651-0000-000, fitted three per pair.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eWill it fit my server?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eLenovo, HPE and Cisco each list a 2-way H200 NVL bridge for particular servers: Lenovo 4X67A97320 for the ThinkSystem SR675 V3 and SR650 V4\/SR650a V4, HPE S4A90C for the ProLiant DL380a Gen12 and DL385 Gen11, and Cisco UCSX-NVL2-H200 for the UCS X580p PCIe Node. Cisco's part is this NVIDIA part, 900-23945-0000-000. Lenovo and HPE qualify their own part numbers, so check with them before fitting this NVIDIA-branded bridge. The two cards must be in adjacent slots, with at least 2.5mm of clearance above their top edge and at least 2.67mm behind the rear card's PCB.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eDoes the bridge combine two cards into one GPU?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eNo. Each H200 NVL keeps its own 141GB of memory and appears as a separate GPU. Software that splits a model or job across both cards, such as tensor-parallel inference or multi-GPU training, can move data between them over NVLink instead of PCIe.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eCan I order it with H200 NVL cards?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eYes. The bridge can be ordered on its own from this page. For two or more H200 NVL cards, use the enquiry link on the H200 NVL page so we can confirm the order with you.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eManufacturer documentation\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cul\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/www.nvidia.com\/en-us\/data-center\/h200\/\"\u003eNVIDIA H200 page, including H200 NVL specifications\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/developer.nvidia.com\/blog\/deploying-nvidia-h200-nvl-at-scale-with-new-enterprise-reference-architecture\/\"\u003eNVIDIA Technical Blog: deploying H200 NVL, NVLink bridge options\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/lenovopress.lenovo.com\/lp1944-nvidia-h200-141gb-gpu\"\u003eLenovo Press: ThinkSystem NVIDIA H200 141GB GPUs product guide\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/pubs.lenovo.com\/sr675-v3\/install_a_gpu_adapter_link_bridge\"\u003eLenovo: installing a GPU link bridge (SR675 V3)\u003c\/a\u003e\u003c\/li\u003e\n\u003cli\u003e\u003ca href=\"https:\/\/buy.hpe.com\/uk\/en\/options\/server-accelerators\/computational-and-graphics-accelerators-for-servers\/nvidia-2%E2%80%91way-nvlink-bridge-for-h200-nvl\/p\/s4a90c\"\u003eHPE: NVIDIA 2-way NVLink bridge for H200 NVL (S4A90C)\u003c\/a\u003e\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003c\/section\u003e","brand":"NVIDIA","offers":[{"title":"Default Title","offer_id":60995486810446,"sku":"900-23945-0000-000","price":495.0,"currency_code":"GBP","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/sr675v3-402-gpu-350.png?v=1790671132","url":"https:\/\/shop.perceptasolutions.com\/products\/nvidia-nvlink-bridge-2-way-for-h200-nvl","provider":"Percepta Solutions","version":"1.0","type":"link"}