{"id":12472,"date":"2026-08-18T04:52:49","date_gmt":"2026-08-18T04:52:49","guid":{"rendered":"https:\/\/rdp.in\/gpu-mart\/product\/nvidia-dgx-spark-personal-ai-supercomputer\/"},"modified":"2026-08-18T04:52:49","modified_gmt":"2026-08-18T04:52:49","slug":"nvidia-dgx-spark-personal-ai-supercomputer","status":"publish","type":"product","link":"https:\/\/rdp.in\/gpu-mart\/product\/nvidia-dgx-spark-personal-ai-supercomputer\/","title":{"rendered":"NVIDIA DGX Spark Personal AI Supercomputer"},"content":{"rendered":"

A petaflop of AI on your desk \u2014 no data centre, no cloud bill, no queue.<\/strong> The NVIDIA DGX Spark puts the Grace Blackwell architecture in a 150 mm box you can carry in one hand, with 128 GB of unified memory<\/strong> that both the CPU and GPU address at once. It runs models locally that normally demand a server \u2014 and it is the most accessible way into serious AI development that RDP sells.<\/p>\n

Key highlights<\/h3>\n
    \n
  • NVIDIA GB10 Grace Blackwell Superchip<\/strong> \u2014 up to 1 petaFLOP of AI performance at FP4<\/strong>.<\/li>\n
  • 128 GB unified LPDDR5x<\/strong> \u2014 CPU and GPU share the full pool over NVLink-C2C<\/strong> at ~5\u00d7 PCIe Gen5 bandwidth. No copying tensors across a bus.<\/li>\n
  • Runs up to ~200B-parameter models locally<\/strong> for inference; fine-tune up to ~70B \u2014 work that normally needs rented cluster time.<\/li>\n
  • Blackwell GPU<\/strong> with 5th-gen Tensor Cores + RT Cores, paired with a 20-core Arm Grace CPU<\/strong>.<\/li>\n
  • 4 TB NVMe<\/strong> onboard for datasets, checkpoints and model weights.<\/li>\n
  • ConnectX-7 200GbE<\/strong> \u2014 link two Sparks to reach roughly 405B-parameter<\/strong> class workloads.<\/li>\n
  • 150 \u00d7 150 mm, ~1.2 kg<\/strong> \u2014 desk-side, plugs into a normal wall socket, no rack or facility work.<\/li>\n
  • Ships with NVIDIA DGX OS<\/strong> and the full NVIDIA AI software stack \u2014 CUDA, containers and frameworks ready on power-up.<\/li>\n<\/ul>\n

    AI workload fit<\/h3>\n
      \n
    • Local inference<\/strong> and model serving up to ~200B params.<\/li>\n
    • Fine-tuning<\/strong> \/ LoRA \/ SFT on models up to ~70B.<\/li>\n
    • RAG<\/strong> and retrieval pipelines against private data that never leaves the desk.<\/li>\n
    • Agentic AI<\/strong> development \u2014 build and iterate on agent fleets locally.<\/li>\n
    • Generative AI<\/strong>, computer vision<\/strong> and NLP & speech<\/strong> prototyping.<\/li>\n
    • Not<\/em> for large-scale pretraining \u2014 that is cluster work (see the upgrade path below).<\/li>\n<\/ul>\n

      AI workload positioning<\/h3>\n

      DGX Spark is where an AI programme starts<\/em>. The unified 128 GB pool is the differentiator: memory capacity, not raw FLOPS, is what usually stops a developer running a large model locally, and Spark removes that wall at desk scale. Develop, fine-tune and validate here on the same CUDA stack you will later run in production \u2014 then scale the identical software to an RDP GPU server or rack-scale system without a rewrite.<\/p>\n

      Industry use cases<\/h3>\n
        \n
      • Enterprises & GCCs:<\/strong> give each AI engineer a private development box instead of contended cloud credits.<\/li>\n
      • Research & education:<\/strong> per-lab or per-researcher AI compute within a departmental budget.<\/li>\n
      • BFSI:<\/strong> prototype on sensitive data that must never leave the premises.<\/li>\n
      • Healthcare:<\/strong> local work on patient data under data-residency obligations.<\/li>\n
      • Media & entertainment:<\/strong> desk-side generative image, video and audio iteration.<\/li>\n<\/ul>\n

        Performance & how to be sure<\/h3>\n

        The headline figure \u2014 1 petaFLOP FP4<\/strong> \u2014 is NVIDIA’s published peak for the GB10 Superchip, and peak is not the same as sustained throughput on your model. Rather than quote a tokens\/sec number that will not match your workload, RDP offers a “benchmark your model”<\/strong> session: bring the model and precision you actually intend to run and we will validate it on this hardware before you commit.<\/p>\n

        Series & upgrade path<\/h3>\n

        Spark is the CARINA<\/strong> (entry) tier of the GPU Mart ladder and the natural first rung. When a single desk-side unit is no longer enough: link a second Spark over ConnectX-7 (~405B class) \u2192 step up to a QUASAR<\/strong> multi-GPU workstation \u2192 an RDP GPU server<\/strong> (8\u00d7 H200 \/ B300) \u2192 rack-scale GB300 NVL72<\/strong>. The software stack is identical at every rung, so nothing is thrown away when you scale.<\/p>\n

        On-prem vs cloud (TCO)<\/h3>\n

        At a fixed capital cost, Spark removes per-hour GPU billing, queueing for scarce instances, and egress charges on your own data \u2014 and keeps everything on-premises for DPDP \/ data-residency purposes. Cloud still wins for burst capacity and very large training runs; Spark wins for the daily, sustained development cycle that would otherwise meter continuously. For a team of engineers billing cloud GPU hours every working day, the payback arithmetic is usually short \u2014 we will model it with your actual usage rather than assert it.<\/p>\n

        Software & day-one readiness<\/h3>\n

        Ships with NVIDIA DGX OS<\/strong> and the NVIDIA AI stack \u2014 CUDA, cuDNN, containers, and the common serving and fine-tuning frameworks (vLLM, PyTorch and the wider NVIDIA NGC catalogue). It is the same CUDA environment as every larger NVIDIA system in this catalogue, which is precisely why work moves upward without a rewrite.<\/p>\n

        Power, thermal & acoustics<\/h3>\n

        Runs from a standard wall socket with an external power adapter \u2014 no rack, no dedicated circuit, no facility cooling<\/strong>. Air-cooled and designed for an office desk rather than a machine room. {Exact wattage, BTU and dB(A) figures: to be published from RDP bench measurement rather than estimated.}<\/em><\/p>\n

        Deployment, warranty & support<\/h3>\n

        In stock and ready to ship.<\/strong> Supplied by RDP Technologies with GST invoice (HSN 8471), pan-India onsite support, and availability through GeM for public-sector procurement. Unbox, plug in, and you are developing the same day.<\/p>\n

        Why RDP<\/h3>\n

        RDP Technologies is a Make-in-India OEM with 14+ years and 300,000+ units shipped, supplying AI infrastructure from desk-side systems to rack-scale AI factories \u2014 with predictable INR pricing, GST invoicing and pan-India onsite service.<\/p>\n

        Buy with confidence<\/h3>\n

        Priced at ₹4,80,000 + applicable taxes<\/strong>. Add to cart, or talk to an RDP solution architect first if you want your model benchmarked on the unit before purchase.<\/p>\n","protected":false},"excerpt":{"rendered":"

        NVIDIA GB10 Grace Blackwell \u00b7 128GB unified LPDDR5x \u00b7 4TB NVMe \u00b7 Desktop<\/p>\n","protected":false},"featured_media":12478,"comment_status":"open","ping_status":"closed","template":"","meta":{"_yoast_wpseo_title":"","_yoast_wpseo_metadesc":"","rank_math_title":"NVIDIA DGX Spark \u2014 Personal AI Supercomputer (GB10, 128GB, 4TB) | RDP GPU Mart","rank_math_description":"Buy the NVIDIA DGX Spark in India: GB10 Grace Blackwell, 1 PFLOP FP4, 128GB unified memory, 4TB NVMe. Runs ~200B-param models on your desk. \u20b94,80,000 + taxes, in stock from RDP.","_hermes_jsonld":""},"product_brand":[],"product_cat":[22],"product_tag":[],"class_list":["post-12472","product","type-product","status-publish","has-post-thumbnail","product_cat-agentic-ai-pcs-edge-ai","pa_form-factor-desktop","pa_gpu-model-nvidia-gb10-grace-blackwell","pa_industry-bfsi-hft","pa_industry-enterprise-gccs","pa_industry-healthcare-life-sciences","pa_industry-media-gaming-entertainment","pa_industry-research-higher-education","pa_series-carina","pa_use-case-agentic-ai","pa_use-case-computer-vision","pa_use-case-fine-tuning","pa_use-case-generative-ai","pa_use-case-inference","pa_use-case-nlp-speech","pa_use-case-rag","first","instock","taxable","shipping-taxable","purchasable","product-type-simple"],"_links":{"self":[{"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/product\/12472","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/product"}],"about":[{"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/types\/product"}],"replies":[{"embeddable":true,"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/comments?post=12472"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/media\/12478"}],"wp:attachment":[{"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/media?parent=12472"}],"wp:term":[{"taxonomy":"product_brand","embeddable":true,"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/product_brand?post=12472"},{"taxonomy":"product_cat","embeddable":true,"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/product_cat?post=12472"},{"taxonomy":"product_tag","embeddable":true,"href":"https:\/\/rdp.in\/gpu-mart\/wp-json\/wp\/v2\/product_tag?post=12472"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}