{"product_id":"r740xd-dual-v100-build-refurbished","title":"R740xd – Dual V100 Build, refurbished","description":"\u003cp dir=\"auto\"\u003e\u003cstrong\u003eDell PowerEdge R740xd 24SFF Rack Server – Dual NVIDIA Tesla V100 Build (Refurbished)\u003c\/strong\u003e\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003e2U Rack Server with Dual NVIDIA Tesla V100 32GB GPUs – Built for AI Training, Deep Learning \u0026amp; Parallel Compute\u003c\/strong\u003e\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003eThis fully configured \u003cstrong\u003eDell PowerEdge R740xd\u003c\/strong\u003e pairs dual Intel Xeon Silver processors with \u003cstrong\u003etwo NVIDIA Tesla V100 32GB GPUs (64GB total GPU memory)\u003c\/strong\u003e, making it a strong choice for multi-GPU AI\/ML training, HPC, and other parallel workloads that outgrow a single-GPU box.\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eKey Specifications:\u003c\/h3\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003cstrong\u003eServer Model\u003c\/strong\u003e: Dell PowerEdge R740xd 24SFF (24x 2.5\" Small Form Factor bays)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eProcessors\u003c\/strong\u003e: 2x Intel Xeon Silver 4210R (2.4GHz base, 10 cores \/ 20 threads each) – 20 Cores \/ 40 Threads total\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eMemory\u003c\/strong\u003e: 128GB DDR4 ECC Registered (expandable)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eGPU\u003c\/strong\u003e: 2x NVIDIA Tesla V100 32GB (64GB total GPU memory), with GPU Enablement Kit\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eStorage\u003c\/strong\u003e: 1x Dell 960GB 6G SATA 2.5\" SSD\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eRAID\/Controller\u003c\/strong\u003e: PERC HBA330 (Non-RAID \/ Pass-Through)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eNetworking\u003c\/strong\u003e: Intel X710 2x 10Gb SFP+ + I350 2x 1Gb Base-T rNDC\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eRemote Management\u003c\/strong\u003e: iDRAC Enterprise (with dedicated management port)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003ePower\u003c\/strong\u003e: 2x 1600W AC Hot-Plug Power Supplies (redundant)\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eCooling\u003c\/strong\u003e: High Performance Fan Kit\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eRails\u003c\/strong\u003e: Included Rail Kit\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eCondition\u003c\/strong\u003e: Professionally refurbished and tested\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eOS\u003c\/strong\u003e: No OS included (supports Windows Server, Linux, VMware ESXi, etc.)\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3 dir=\"auto\"\u003eHighlights:\u003c\/h3\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003cstrong\u003eDual NVIDIA Tesla V100 32GB GPUs\u003c\/strong\u003e – 64GB combined GPU memory for multi-GPU training, large-model inference, and compute-intensive HPC workloads\u003c\/li\u003e\n\u003cli\u003e24-bay 2.5\" SFF chassis with room for storage expansion\u003c\/li\u003e\n\u003cli\u003eStrong dual-socket performance with 20 cores \/ 40 threads\u003c\/li\u003e\n\u003cli\u003eEnterprise remote management via iDRAC Enterprise\u003c\/li\u003e\n\u003cli\u003eHigh-bandwidth 10Gb SFP+ networking included\u003c\/li\u003e\n\u003cli\u003eRedundant 1600W power supplies + high-performance cooling for 24\/7 reliability\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003ePerfect For:\u003c\/strong\u003e\u003c\/p\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003eMulti-GPU AI \/ Machine Learning training\u003c\/li\u003e\n\u003cli\u003eDeep learning research and large-batch inference\u003c\/li\u003e\n\u003cli\u003eHigh Performance Computing (HPC) and scientific computing\u003c\/li\u003e\n\u003cli\u003eGPU-accelerated virtualization and VDI\u003c\/li\u003e\n\u003cli\u003eRendering and 3D graphics workloads\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp dir=\"auto\"\u003eThis server has been fully tested and is ready to deploy. Ships with all necessary power cables, rails, and bezel (where applicable).\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eNote\u003c\/strong\u003e: This is a refurbished unit in excellent cosmetic and functional condition. Minor signs of previous rack use may be present.\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eCompare Our R740xd Builds:\u003c\/h3\u003e\n\u003cp dir=\"auto\"\u003eAll four builds share the same GPU-ready platform (24SFF chassis, 128GB RAM, Intel X710\/I350 networking, iDRAC Enterprise, GPU enablement kit, high-performance fans, 2x 1600W supplies) — they differ in CPU, storage controller and GPU:\u003c\/p\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/products\/r740xd-gpu-ready-no-gpu-included-build-refurbished\"\u003e\u003cstrong\u003eR740xd – GPU Ready (No GPU Included)\u003c\/strong\u003e\u003c\/a\u003e – $2,995 – 2x Silver 4210R, HBA330 pass-through. The lowest-cost GPU-ready host; bring your own card.\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/products\/r740xd-gpu-ready-no-gpu-included-build-refurbished-1\"\u003e\u003cstrong\u003eR740xd – GPU Ready, Gold 6248 \/ H740p\u003c\/strong\u003e\u003c\/a\u003e – $3,493.33 – 2x Gold 6248 (40 cores), PERC H740p RAID. More CPU and hardware RAID; bring your own card.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eR740xd – Dual V100 Build (this listing)\u003c\/strong\u003e – $5,849 – Silver 4210R platform + 2x Tesla V100 32GB (64GB total). Best value for multi-GPU training and parallel workloads.\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/products\/r740xd-a40-build-refurbished\"\u003e\u003cstrong\u003eR740xd – A40 Build\u003c\/strong\u003e\u003c\/a\u003e – $10,959 – Silver 4210R platform + 1x NVIDIA A40 48GB. Best single-GPU performance for AI inference and visualization.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp dir=\"auto\"\u003eNeed a custom configuration, more storage, or a different GPU pairing? Message us — we can build to your requirements.\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eWhat Two V100 32GB Cards Actually Give You\u003c\/h3\u003e\n\u003cp dir=\"auto\"\u003eWorth being precise about the \"64GB\" number, because it is the thing buyers most often misread. The Tesla V100 PCIe has \u003cstrong\u003eno NVLink connector\u003c\/strong\u003e — NVLink was exclusive to the SXM2 module version. These two cards talk to each other over \u003cstrong\u003ePCIe Gen3 x16\u003c\/strong\u003e, not a high-speed GPU-to-GPU link. That means:\u003c\/p\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003cstrong\u003e64GB is two pools of 32GB, not one pool of 64GB.\u003c\/strong\u003e A model that fits in 32GB runs beautifully on one card. A model larger than 32GB has to be sharded across both, and pays PCIe bandwidth on every layer boundary.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eWhere two cards shine:\u003c\/strong\u003e data-parallel training (two independent copies chewing through different batches), running two different models side by side, per-user or per-tenant GPU assignment, batch inference, and traditional CUDA\/HPC work that was written for multi-GPU nodes.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eRaw bandwidth per card is excellent.\u003c\/strong\u003e Each V100 carries 32GB of HBM2 at up to \u003cstrong\u003e900 GB\/s\u003c\/strong\u003e — meaningfully faster than the GDDR6 on newer 48GB cards. For anything that fits inside 32GB, a V100 is still a quick card.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eSpecs per card:\u003c\/strong\u003e 5,120 CUDA cores, 640 Tensor cores, 250W, dual-slot, passively cooled, PCIe Gen3 x16. Two cards means 500W of GPU load in the chassis.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3 dir=\"auto\"\u003eSoftware Notes Before You Buy\u003c\/h3\u003e\n\u003cp dir=\"auto\"\u003eThe V100 is Volta (compute capability sm70), and that has real consequences for modern LLM tooling that are easy to discover the hard way:\u003c\/p\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003cstrong\u003eFP16 tensor cores, but no BF16 and no FP8.\u003c\/strong\u003e Volta predates both. Training recipes written for BF16 need adapting to FP16 with loss scaling.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eSeveral modern inference kernels target Ampere (sm80) and newer.\u003c\/strong\u003e FlashAttention-2 and the Marlin INT4 kernels are the common examples; on Volta you run FP16 paths, older kernel implementations, or community sm70 builds of vLLM. It works — people do run quantized models on V100s — but expect to spend time on the stack rather than pulling a container and going.\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eExcellent for FP16 training, CUDA and HPC.\u003c\/strong\u003e Simulation, scientific computing, rendering, classic deep-learning training and CUDA-native codebases all run without ceremony.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp dir=\"auto\"\u003ePut plainly: if the plan is serving modern INT4-quantized language models with off-the-shelf tooling, our \u003ca href=\"https:\/\/resilient-tec.com\/products\/r740xd-a40-build-refurbished\"\u003eA40 build\u003c\/a\u003e is the better fit despite the higher price — 48GB in a single Ampere pool with full modern kernel support. If the plan is FP16 training, multi-tenant GPU workloads, HPC, or rendering, two V100s at this price is hard to beat. Not sure which describes your workload? \u003ca href=\"https:\/\/resilient-tec.com\/pages\/request-a-quote\"\u003eTell us what you intend to run\u003c\/a\u003e and we will say so honestly.\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eConfiguration Requirements — Already Handled\u003c\/h3\u003e\n\u003cp dir=\"auto\"\u003eBoth V100s are 250W \u003cstrong\u003epassively cooled\u003c\/strong\u003e cards with no fans of their own. Dropping them into a stock R740xd does not work — Dell's installation guidance requires a specific set of parts, all of which are installed and tested in this build:\u003c\/p\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003cstrong\u003eGPU enablement kit\u003c\/strong\u003e (risers, GPU power cables, GPU air shroud, mylar insulation) — Dell 330-BBMO \/ 85CYN in the US, 490-BEIX \/ JKFGX in EMEA\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eLow-profile heat sinks in both CPU sockets\u003c\/strong\u003e — standard heat sinks physically foul the GPU air shroud, and Dell requires both sockets populated\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eHigh-performance fan kit\u003c\/strong\u003e — mandatory for any GPU configuration in this chassis, and doubly so with 500W of passive GPU in the air path\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003eDedicated GPU power cables\u003c\/strong\u003e — required for any card above 75W; each V100 draws 250W\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003e1600W power supplies\u003c\/strong\u003e — Dell sets an 1100W floor for GPU configurations; 1600W supplies provide real redundancy headroom rather than a nominal pair\u003c\/li\u003e\n\u003cli\u003e\n\u003cstrong\u003e30°C maximum inlet temperature\u003c\/strong\u003e — the thermal ceiling once GPUs are installed alongside 150W\/8C, 165W\/12C, 200W or 205W processors\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003cp dir=\"auto\"\u003eAssembled from separate secondary-market purchases, that is seven line items, seven condition claims and seven return windows. Here it is one part number, one warranty and one place to call. The full breakdown, including the 330-BBMO vs 490-BEIX part-number confusion and the two mistakes that stop most self-assembled builds: \u003ca href=\"https:\/\/resilient-tec.com\/blogs\/news\/nvidia-a40-poweredge-r740xd-gpu-enablement-kit-requirements\"\u003e\u003cstrong\u003ePutting a passive datacenter GPU in a PowerEdge R740xd\u003c\/strong\u003e\u003c\/a\u003e\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eFrequently Asked Questions\u003c\/h3\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eIs this really 64GB of usable GPU memory?\u003c\/strong\u003e\u003cbr\u003eIt is 2 x 32GB, not a single 64GB pool. The Tesla V100 PCIe has no NVLink connector, so the cards communicate over PCIe Gen3 x16. Models that fit in 32GB run on one card at full speed; larger models must be sharded across both and pay PCIe bandwidth at each layer boundary.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eCan I run a 70B language model on this server?\u003c\/strong\u003e\u003cbr\u003eYes, sharded across both cards at 4-bit quantization, but expect to invest effort in the software stack — several modern INT4 and attention kernels target Ampere and newer. For straightforward 70B serving, our A40 build holds the model in one 48GB pool with full modern kernel support.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eDoes the V100 support BF16 or FP8?\u003c\/strong\u003e\u003cbr\u003eNeither. Volta tensor cores are FP16. BF16 arrived with Ampere and FP8 with Ada Lovelace and Hopper. FP16 training with loss scaling works well.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eWhat is the V100's memory bandwidth compared to newer cards?\u003c\/strong\u003e\u003cbr\u003eUp to 900 GB\/s of HBM2 per card, which is higher than the 696 GB\/s GDDR6 on a 48GB A40. For workloads that fit inside 32GB, the V100 is genuinely fast.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eCan I add a third GPU?\u003c\/strong\u003e\u003cbr\u003eThe R740xd supports up to three double-width accelerators, and Dell requires all installed GPUs to be identical in type and model — so a third V100, not a mixed pair. Contact us first; power and riser assignments change.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eDoes it come with an operating system?\u003c\/strong\u003e\u003cbr\u003eNo OS is included. The server supports Windows Server, Linux distributions, VMware ESXi and Proxmox. iDRAC Enterprise is licensed for remote installation and management.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eCan I expand the storage?\u003c\/strong\u003e\u003cbr\u003eYes. The chassis has 24 x 2.5\" SFF bays and ships with one 960GB SATA SSD, leaving 23 bays open. The PERC HBA330 runs in pass-through mode, which suits ZFS, Ceph and software-defined storage.\u003c\/p\u003e\n\u003cp dir=\"auto\"\u003e\u003cstrong\u003eWhat condition is it in and what warranty applies?\u003c\/strong\u003e\u003cbr\u003eProfessionally refurbished, fully tested and burned in, in excellent cosmetic and functional condition with minor signs of previous rack use possible. Backed by the Resilient Tec lifetime warranty, reachable directly rather than through a manufacturer ticket queue.\u003c\/p\u003e\n\u003ch3 dir=\"auto\"\u003eFurther Reading\u003c\/h3\u003e\n\u003cul dir=\"auto\"\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/blogs\/news\/nvidia-a40-poweredge-r740xd-gpu-enablement-kit-requirements\"\u003ePutting a passive datacenter GPU in a PowerEdge R740xd\u003c\/a\u003e — every part the cards need, the enablement kit part numbers, and why a stock chassis throttles them\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/blogs\/news\/what-can-you-run-on-48gb-nvidia-a40-llm-model-guide-2026\"\u003eWhat can you actually run on 48GB of GPU memory?\u003c\/a\u003e — benchmarked model sizes and the KV cache math, useful for sizing either build\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/blogs\/news\/custom-dell-r740xd-gpu-server-builds-prices-faq\"\u003eCustom Dell R740xd GPU servers: build options, prices, and the forum questions nobody answers\u003c\/a\u003e — which GPU power cable (4VPD3 vs TR5TP), fans at 100%, LFF chassis limits, and all four builds compared\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/blogs\/news\/the-cloud-sent-you-a-bill-time-to-come-home\"\u003eThe cloud sent you a bill\u003c\/a\u003e — on-prem compute versus rented GPU time\u003c\/li\u003e\n\u003cli\u003e\n\u003ca href=\"https:\/\/resilient-tec.com\/pages\/gpu-ai-infrastructure-guide\"\u003eGPU \u0026amp; AI infrastructure guide\u003c\/a\u003e and \u003ca href=\"https:\/\/resilient-tec.com\/pages\/enterprise-servers-guide\"\u003eenterprise servers guide\u003c\/a\u003e\n\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003cscript type=\"application\/ld+json\"\u003e\n{\n  \"@context\": \"https:\/\/schema.org\",\n  \"@type\": \"FAQPage\",\n  \"mainEntity\": [\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Do two Tesla V100 32GB cards give 64GB of usable GPU memory?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"It is 2 x 32 GB, not a single 64 GB pool. The Tesla V100 PCIe has no NVLink connector, so the cards communicate over PCIe Gen3 x16. Models that fit in 32 GB run on one card at full speed; larger models must be sharded across both and pay PCIe bandwidth at each layer boundary.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Can a dual Tesla V100 server run a 70B language model?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"Yes, sharded across both cards at 4-bit quantization, but expect to invest effort in the software stack because several modern INT4 and attention kernels target Ampere and newer. For straightforward 70B serving, a single 48GB Ampere card holds the model in one pool with full modern kernel support.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Does the NVIDIA Tesla V100 support BF16 or FP8?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"Neither. Volta tensor cores are FP16. BF16 arrived with Ampere and FP8 with Ada Lovelace and Hopper. FP16 training with loss scaling works well on the V100.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"What is the Tesla V100's memory bandwidth?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"Up to 900 GB\/s of HBM2 per card, which is higher than the 696 GB\/s GDDR6 on a 48GB A40. For workloads that fit inside 32 GB, the V100 is genuinely fast.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Can a third GPU be added to this R740xd?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"The R740xd supports up to three double-width accelerators, and Dell requires all installed GPUs to be identical in type and model, so a third V100 rather than a mixed pair. Power and riser assignments change, so check before ordering.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Does this server come with an operating system?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"No operating system is included. The server supports Windows Server, Linux distributions, VMware ESXi and Proxmox. iDRAC Enterprise is licensed for remote installation and management.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"Can the storage be expanded?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"Yes. The chassis has 24 x 2.5-inch SFF bays and ships with one 960GB SATA SSD, leaving 23 bays open. The PERC HBA330 runs in pass-through mode, which suits ZFS, Ceph and software-defined storage.\" }\n    },\n    {\n      \"@type\": \"Question\",\n      \"name\": \"What condition is this server in and what warranty applies?\",\n      \"acceptedAnswer\": { \"@type\": \"Answer\", \"text\": \"Professionally refurbished, fully tested and burned in, in excellent cosmetic and functional condition with minor signs of previous rack use possible. Backed by the Resilient Tec lifetime warranty, reachable directly rather than through a manufacturer ticket queue.\" }\n    }\n  ]\n}\n\u003c\/script\u003e","brand":"Dell","offers":[{"title":"Default Title","offer_id":48070124830881,"sku":"R740XD-V100-R","price":175470.0,"currency_code":"BRL","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/2216\/8673\/files\/R740xd-Dual-V100-Build-refurbished.jpg?v=1791961854","url":"https:\/\/www.foxti.com.br\/products\/r740xd-dual-v100-build-refurbished","provider":"FoxTI","version":"1.0","type":"link"}