{"product_id":"geforce-rtx-4090-48gb-gddr6x-blower-graphics-card-modified","title":"GeForce RTX 4090 48GB GDDR6X Blower Graphics Card - Modified","description":"\u003csection class=\"phd-intro\"\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003eLocal AI \/ Modified 48GB workstation GPU\u003c\/p\u003e\n\u003ch2\u003eMore room for local AI.\u003c\/h2\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cp\u003eBuild your local AI workstation around 48GB of dedicated GDDR6X memory. This modified GeForce RTX 4090 configuration is aimed at running language models, developing local AI applications and experimenting with memory-intensive workloads on your own hardware. Active blower cooling and a dual-slot design support workstation integration; the exact model, settings and software determine what fits.\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\"\u003eWhat's included\u003c\/a\u003e\u003ca href=\"#hardware-documents\"\u003eTechnical documents\u003c\/a\u003e\u003c\/nav\u003e\n\u003cdiv class=\"phd-metrics\"\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e48GB\u003c\/strong\u003e\u003cspan\u003eGDDR6X memory per card\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e16,384\u003c\/strong\u003e\u003cspan\u003eCUDA cores\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e384-bit\u003c\/strong\u003e\u003cspan\u003eMemory interface\u003c\/span\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cstrong\u003e38.5mm\u003c\/strong\u003e\u003cspan\u003eNominal card thickness\u003c\/span\u003e\n\u003c\/div\u003e\n\u003c\/div\u003e\n\u003caside class=\"phd-note\"\u003e\n\u003ch3\u003eModified configuration, not a standard NVIDIA 48GB model\u003c\/h3\u003e\n\u003cp\u003eThe standard GeForce RTX 4090 has 24GB of memory. This listing is for an independently produced 48GB board configuration based on the RTX 4090 GPU, not a card manufactured by NVIDIA. It is sold as New and covered by Percepta's 12-month return-to-base warranty for independently modified graphics cards. The larger memory capacity does not imply RTX professional-card certification, pooled multi-GPU memory or NVIDIA manufacturer warranty cover.\u003c\/p\u003e\n\u003c\/aside\u003e\n\u003csection class=\"phd-components\" id=\"hardware-components\"\u003e\n\u003ch2\u003eA closer look at the card\u003c\/h2\u003e\n\u003cdiv class=\"phd-features\"\u003e\n\u003carticle\u003e\n\u003cfigure class=\"phd-figure\"\u003e\u003cimg src=\"https:\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/rtx4090-48gb-hero.png?v=1790374024\" alt=\"Modified GeForce RTX 4090 48GB card, oblique view showing blower housing and PCIe edge connector\" loading=\"lazy\"\u003e\n\u003cfigcaption\u003eOriginal card photograph \/ oblique view\u003c\/figcaption\u003e\n\u003c\/figure\u003e\n\u003ch3\u003e48GB on a single card\u003c\/h3\u003e\n\u003cp\u003eKeep more of a local AI workload in GPU memory: model weights, context and runtime working space. The 48GB configuration offers extra capacity for memory-intensive inference, with the exact model, quantisation and settings determining what fits.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003carticle\u003e\n\u003cfigure class=\"phd-figure\"\u003e\u003cimg src=\"https:\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/rtx4090-48gb-front.png?v=1790374024\" alt=\"Front of the 48GB RTX 4090 showing its radial blower fan and full-length shroud\" loading=\"lazy\"\u003e\n\u003cfigcaption\u003eOriginal card photograph \/ blower and shroud\u003c\/figcaption\u003e\n\u003c\/figure\u003e\n\u003ch3\u003eActive blower cooling\u003c\/h3\u003e\n\u003cp\u003eThe enclosed shroud and radial fan direct air towards the vented rear bracket. The nominal 266.7mm length and 38.5mm thickness help with planning a workstation installation. Allow clearance for the fan inlet, cabling and adjacent cards.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003carticle\u003e\n\u003cfigure class=\"phd-figure\"\u003e\u003cimg src=\"https:\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/rtx4090-48gb-rear.png?v=1790374024\" alt=\"Rear PCB and mounting plate of the modified RTX 4090 48GB card\" loading=\"lazy\"\u003e\n\u003cfigcaption\u003eOriginal card photograph \/ rear PCB and mounting\u003c\/figcaption\u003e\n\u003c\/figure\u003e\n\u003ch3\u003ePCIe workstation integration\u003c\/h3\u003e\n\u003cp\u003eA PCIe 4.0 x16 host interface connects the card to a compatible system. The board provides three DisplayPort 1.4a outputs and one HDMI 2.1 output. Check the complete host configuration, power connection and driver requirements before installation.\u003c\/p\u003e\n\u003c\/article\u003e\n\u003c\/div\u003e\n\u003c\/section\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\u003eFigures below describe the supplier's 48GB configuration. They are not a specification for the standard 24GB retail card.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003ctable\u003e\n\u003ccaption\u003e01 \/ Graphics processor and memory\u003c\/caption\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eGPU\u003c\/th\u003e\n\u003ctd\u003eNVIDIA AD102-301-A1 \/ GeForce RTX 4090-based\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eArchitecture\u003c\/th\u003e\n\u003ctd\u003eNVIDIA Ada Lovelace\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCUDA cores\u003c\/th\u003e\n\u003ctd\u003e16,384\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eGPU memory\u003c\/th\u003e\n\u003ctd\u003e48GB GDDR6X (49,152MB reported in the supplied GPU-Z record)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eMemory interface\u003c\/th\u003e\n\u003ctd\u003e384-bit\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eMemory data rate\u003c\/th\u003e\n\u003ctd\u003e21Gbps stated\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eMemory bandwidth\u003c\/th\u003e\n\u003ctd\u003e1,008.4GB\/s reported in the supplied GPU-Z record\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eBase \/ boost clock\u003c\/th\u003e\n\u003ctd\u003e2,235MHz \/ 2,520MHz stated; operating clocks vary with conditions\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003ctable\u003e\n\u003ccaption\u003e02 \/ Interface, cooling and physical layout\u003c\/caption\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eHost interface\u003c\/th\u003e\n\u003ctd\u003ePCIe 4.0 x16\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eDisplay connections\u003c\/th\u003e\n\u003ctd\u003e3 x DisplayPort 1.4a; 1 x HDMI 2.1\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eCooling\u003c\/th\u003e\n\u003ctd\u003eActive radial blower with vented rear bracket\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eStated TDP\u003c\/th\u003e\n\u003ctd\u003e450W; not a measurement of complete-system consumption\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eNominal dimensions\u003c\/th\u003e\n\u003ctd\u003e266.7 x 111 x 38.5mm (length x height x thickness)\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eForm factor\u003c\/th\u003e\n\u003ctd\u003eFull-height, dual-slot\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003ePCB\u003c\/th\u003e\n\u003ctd\u003e14 layers, matt-black finish\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr\u003e\n\u003cth scope=\"row\"\u003eSoftware interfaces listed\u003c\/th\u003e\n\u003ctd\u003eCUDA, DirectX 12 Ultimate, OpenGL 4.6 and Vulkan; validate your required driver and application combination\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\u003cp class=\"phd-footnote\"\u003eBased on supplier specification BP49K-EA46B and the accompanying hardware record. Exact power-connector requirements, supplied accessories and compatibility with a particular host must be confirmed before ordering. Supplier test results are reference evidence, not a performance guarantee for your system.\u003c\/p\u003e\n\u003c\/section\u003e\n\u003csection class=\"phd-supply\" id=\"hardware-included\"\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003eThis listing\u003c\/p\u003e\n\u003ch2\u003eCard, not a complete system\u003c\/h2\u003e\n\u003cul\u003e\n\u003cli\u003eOne GeForce RTX 4090-based 48GB blower graphics card is the proposed configuration.\u003c\/li\u003e\n\u003cli\u003ePackaging, cables, adapters and other accessories are not yet confirmed.\u003c\/li\u003e\n\u003cli\u003eNo software licence or complete workstation is represented as included.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003c\/div\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003ePlan your setup\u003c\/p\u003e\n\u003ch2\u003eWhat you'll need\u003c\/h2\u003e\n\u003cul\u003e\n\u003cli\u003eA compatible PCIe host with full-height, dual-slot clearance.\u003c\/li\u003e\n\u003cli\u003eA suitable power supply and the correct GPU power cable, confirmed for this board.\u003c\/li\u003e\n\u003cli\u003eClear blower intake and exhaust paths, with room for power-cable routing.\u003c\/li\u003e\n\u003cli\u003eA validated operating system, driver and application stack for the modified memory configuration.\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\u003ch2\u003ePlanning your local AI setup\u003c\/h2\u003e\n\u003cdetails\u003e\n\u003csummary\u003eWhy choose 48GB for local AI?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eVRAM holds the model and the working memory it needs while generating an answer. Compared with a 24GB card, 48GB gives you more room for model weights, longer conversations and larger workloads. The benefit is capacity: twice the memory does not mean twice the speed.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eWhat size language models can I run?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eCheck the exact model and quantisation, not just its parameter count. Lower-bit versions, such as 4-bit or 8-bit, reduce the space needed for model weights. Leave room for the context cache and runtime overhead too: longer prompts and parallel requests increase memory use. A model file smaller than 48GB is not, by itself, proof that the complete workload will fit.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eCan I use Ollama or LM Studio?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eThese are common starting points for running local language models. \u003ca href=\"https:\/\/docs.ollama.com\/gpu\" rel=\"noopener\" target=\"_blank\"\u003eOllama lists the RTX 4090 among its supported NVIDIA GPUs\u003c\/a\u003e, but that is not validation of this modified 48GB board. Confirm the driver, operating system and chosen runtime for the supplied card before committing to a deployment.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eCan my models and documents stay on my own machine?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eYes, with a locally configured AI application. For example, \u003ca href=\"https:\/\/lmstudio.ai\/docs\/app\/offline\" rel=\"noopener\" target=\"_blank\"\u003eLM Studio supports offline inference\u003c\/a\u003e after the model and runtime have been downloaded. Use local models and local document processing, and check cloud features, external tools and integrations separately. Owning the GPU alone does not make every application private or offline.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eIs this for inference, or can I fine-tune models too?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eLocal inference is the main focus: running an existing model on your own workstation. Memory-efficient adaptation methods such as LoRA and QLoRA may also be suitable, depending on the model, sequence length, batch size and software. \u003ca href=\"https:\/\/huggingface.co\/docs\/transformers\/quantization\/bitsandbytes\" rel=\"noopener\" target=\"_blank\"\u003eQLoRA uses quantised base weights with trainable adapters\u003c\/a\u003e; it is different from full-model training. No specific fine-tuning configuration has been validated for this listing.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003cdetails\u003e\n\u003csummary\u003eWould two cards give me 96GB for one model?\u003c\/summary\u003e\n\u003cdiv\u003e\n\u003cp\u003eTwo cards provide 48GB each, not one automatic 96GB memory pool. A compatible inference engine can split a model across GPUs; \u003ca href=\"https:\/\/docs.ollama.com\/faq#how-does-ollama-load-models-on-multiple-gpus\" rel=\"noopener\" target=\"_blank\"\u003eOllama documents this approach\u003c\/a\u003e when a model does not fit on one GPU. Your workstation still needs the PCIe slots, power, cooling and software support for both cards.\u003c\/p\u003e\n\u003c\/div\u003e\n\u003c\/details\u003e\n\u003c\/section\u003e\n\u003csection class=\"phd-contact phd-specs\" id=\"hardware-documents\"\u003e\n\u003cdiv\u003e\n\u003cp class=\"phd-eyebrow\"\u003eTechnical documents\u003c\/p\u003e\n\u003ch2\u003eRTX 4090 48GB data sheet\u003c\/h2\u003e\n\u003cp\u003eSpecifications, dimensions and original product photographs.\u003c\/p\u003e\n\u003cp class=\"phd-eyebrow\"\u003ePDF \/ 5 pages \/ 2.97 MB\u003c\/p\u003e\n\u003c\/div\u003e\n\u003ca class=\"button button--secondary\" href=\"https:\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/Percepta_RTX4090_48GB_Data_Sheet.pdf?v=1790374075\" rel=\"noopener\" target=\"_blank\"\u003eDownload PDF \u003cspan\u003e  ↓\u003c\/span\u003e\u003c\/a\u003e\u003c\/section\u003e","brand":"Percepta Solutions","offers":[{"title":"Default Title","offer_id":60957431103822,"sku":null,"price":4195.0,"currency_code":"GBP","in_stock":false}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1078\/0213\/2814\/files\/rtx4090-48gb-hero.png?v=1790374024","url":"https:\/\/shop.perceptasolutions.com\/products\/geforce-rtx-4090-48gb-gddr6x-blower-graphics-card-modified","provider":"Percepta Solutions","version":"1.0","type":"link"}