{"title":"NVIDIA DGX Spark","description":"\u003cp\u003eDer NVIDIA DGX Spark ist ein kompakter KI-Supercomputer, verfügbar von mehreren OEM-Herstellern, sowie in einer Founders Edition direkt von NVIDIA.  Eine großzügige 128GB RAM-Ausstattung ermöglicht die Arbeit mit großen LLMs mit bis zu 200B Parametern.\u003c\/p\u003e\n\u003cp\u003eKern des NVIDIA DGX Spark ist der NVIDIA® GB10 Grace Blackwell Superchip, mit einer Leistung von 1 Petaflop für KI-Rechnungen (in FP4-Precision).\u003c\/p\u003e\n\u003cp\u003eDie NVIDIA DGX Spark Plattform ist besonders für AI-Entwickler und KI-Forschung geeignet. Dank des kompakten Formfaktors findet der NVIDIA DGX Spark auf jedem Schreibtisch Platz.\u003c\/p\u003e\n\u003cp\u003eWesentlicher Teil der Plattform ist der NVIDIA AI Software Stack, der einen schnellen Einstieg mit der am besten unterstützten Plattform in der LLM-Industrie ermöglicht.  Dank der Software-Kompatibilität mit NVIDIA's größeren Plattformen (bspw. H200 \/ B200) können die entwickelten LLM-Lösungen und Anwendungen (bspw. agentic AI) direkt und problemlos skaliert werden. \u003c\/p\u003e\n\u003cp\u003eDank des eingebauten NVIDIA® ConnectX-7 Netzwerklinks können zwei NVIDIA DGX Spark Systeme miteinander verbunden werden, um noch größere Modelle bis zu 405B Parametern zu unterstützen. \u003cbr\u003e\u003c\/p\u003e\n\u003cp\u003eFür höhere Anforderungen im Office-Umfeld stellt \u003ca href=\"https:\/\/buyzero.de\/collections\/nvidia-dgx-station-gb300\"\u003eNVIDIA die DGX Station GB300\u003c\/a\u003e bereit.\u003c\/p\u003e","products":[{"product_id":"asus-ascent-gx10-nvidia-dgx-spark","title":"ASUS Ascent GX10 - NVIDIA DGX Spark 1\/4TB","description":"\u003ch1 dir=\"ltr\"\u003e \u003cspan\u003eASUS Ascent GX10: AI supercomputer based on the NVIDIA DGX Spark platform\u003c\/span\u003e\n\u003c\/h1\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eThe ASUS Ascent GX10 is a compact AI supercomputer. A generous 128GB of RAM enables work with large LLMs with up to 200B parameters.\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eThe core of the ASUS Ascent GX10 is the \u003cspan class=\"fontstyle0\"\u003eNVIDIA® GB10 Grace Blackwell superchip, with a performance of 1 petaflop for AI calculations (in FP4 precision).\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eThe NVIDIA DGX Spark platform is especially well-suited for AI developers and AI research. Thanks to its compact form factor, the ASUS Ascent GX10 fits perfectly on any desktop.\u003c\/span\u003e\u003c\/p\u003e\n\n \u003cp dir=\"ltr\"\u003e\u003cspan\u003eA key component of the platform is the NVIDIA AI software stack, which enables a quick start with the best-supported platform in the LLM industry. Thanks to software compatibility with NVIDIA's larger platforms (e.g., H200 \/ B200), the developed LLM solutions and applications (e.g., agentic AI) can be scaled directly and easily.\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eThanks to the built-in NVIDIA® ConnectX-7 network link, two NVIDIA DGX Spark systems can be connected to support even larger models up to 405B parameters.\u003cbr\u003e\u003cbr\u003e \u003cstrong\u003e\u003cspan style=\"color: rgb(0, 0, 0);\"\u003eDelivery time\u003c\/span\u003e\u003c\/strong\u003e\u003cbr data-start=\"121\" data-end=\"124\"\u003e \u003cspan style=\"color: rgb(0, 0, 0);\"\u003eDelivery for the ASUS Ascent GX10 – NVIDIA DGX Spark typically takes \u003cstrong\u003e4–6 weeks from receipt of order.\u003c\/strong\u003e Please take this into account when planning your project.\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp\u003e\u003c\/p\u003e\n\n\u003ch2 dir=\"ltr\"\u003e \u003cspan\u003eTechnical data\u003c\/span\u003e \u003c\/h2\u003e\n\n\u003cdiv dir=\"ltr\" align=\"left\"\u003e\n\n\u003ctable style=\"width: 100%; height: 498.313px;\"\u003e\n\n\u003ccolgroup\u003e\n\u003ccol width=\"139\" style=\"width: 28.9991%;\"\u003e\n\u003ccol width=\"421\" style=\"width: 70.6385%;\"\u003e\n\u003c\/colgroup\u003e\n\n\u003ctbody\u003e\n\n\u003ctr style=\"height: 35.5938px;\"\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cstrong\u003efeature\u003c\/strong\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cstrong\u003especification\u003c\/strong\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5938px;\"\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eAI computing power\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n \u003cp dir=\"ltr\"\u003e\u003cspan\u003e1 petaflops (FP4, sparse) (NVIDIA Grace Blackwell)\u003c\/span\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5938px;\"\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eR.A.M.\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e128GB RAM LPDDR5x (Unified memory)\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eRAM bandwidth\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan role=\"presentation\" dir=\"ltr\"\u003e273 GB\/s (256-bit memory interface)\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eGPU\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan role=\"presentation\" dir=\"ltr\"\u003eNVIDIA Blackwell Architecture\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eCPU\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan role=\"presentation\" dir=\"ltr\"\u003e20 core ARM: 10 Cortex-X925\u003c\/span\u003e \u003cspan role=\"presentation\" dir=\"ltr\"\u003e+ 10 Cortex-A725 ARM\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003ememory\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x M.2 2242\/2230 PCIe Gen5x4 (1TB\/2TB\/4TB Value \u0026amp; Performance), PCIe Gen4 compatible\u003cbr\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eWi-Fi \u0026amp; Bluetooth\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n \u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eAW-EM637 Wi-Fi 7 (Gig+) 2x2 + Bluetooth® 5.3 with Bluetooth LE\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003enetwork\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x 10 GBit LAN\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x ConnectX CX-7 Smart NIC (200 Gbps, 2 x QSFP)\u003cbr\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eUSB ports\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e3 x USB 3.2 Gen 2x2 Type-C, 20Gbps, alternate mode (DisplayPort 2.1)\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x USB 3.2 Gen 2x2 Type-C, 20 Gbps, with PD in(180W EPR PD3.1 SPEC)\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eAdditional ports \u0026amp; features\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x HDMI 2.1\u003cbr\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x Kensington Lock\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x on\/off button\u003c\/span\u003e\u003c\/span\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5938px;\"\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eOperating system\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eNVIDIA DGX™ OS (Ubuntu Linux)\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5938px;\"\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eDimensions\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5938px;\"\u003e\n\n \u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e150 x 150 x 51 mm\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003eWeight\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003cspan\u003e1.48 kg\u003c\/span\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\n\u003c\/tbody\u003e\n\n\n\u003c\/table\u003e\n\n\n\u003c\/div\u003e\n\n\u003ch2 dir=\"ltr\"\u003e\u003cspan\u003eScope of delivery\u003c\/span\u003e\u003c\/h2\u003e\n\n\u003cul\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e \u003cspan\u003e1 × ASUS Ascent GX 10 (based on NVIDIA DGX Spark)\u003c\/span\u003e \u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e\u003cspan\u003e\u003c\/span\u003e power supply\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e Power cable (EU)\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e \u003cspan\u003eNote: QSFP cable for the ConnectX CX-7 network interface, for connecting two NVIDIA DGX Spark systems, is not included - it is sold separately\u003c\/span\u003e\n\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\n\u003ch2 dir=\"ltr\"\u003e Areas of application\u003c\/h2\u003e\n\n\u003cul\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e \u003cspan\u003eIndustry \u0026amp; Research\u003c\/span\u003e \u003cspan\u003e: Development and fine-tuning of AI LLM models with up to 200B parameters (or 405B when using two ASUS Ascent GX10 \/ DGX Sparks)\u003c\/span\u003e\u003cspan\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e Universities \u0026amp; Colleges: Experiments and hands-on experience for students at large LLM models\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli dir=\"ltr\" aria-level=\"1\"\u003e\n\n\u003cp dir=\"ltr\" role=\"presentation\"\u003e Offices, business environment: Running Agentic AI, especially for data protection - e.g. for processing customer inquiries, offline email processing, RAGs with sensitive company data.\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\n\u003ch2 dir=\"ltr\"\u003e \u003cspan\u003eNVIDIA DGX Spark product variants\u003c\/span\u003e\n\u003c\/h2\u003e\n\n \u003cp dir=\"ltr\"\u003eThe NVIDIA DGX Spark platform is built on NVIDIA components at its core. Every variant features the Grace Blackwell GB10 superchip and 128 GB of RAM with 273 GB\/s memory bandwidth. Memory bandwidth has the greatest impact on token\/s throughput rates.\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e The operating system (NVIDIA DGX OS, based on Ubuntu) and the NVIDIA AI software stack are the same in all cases.\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e Various providers build a housing around it and sell the device under their own brand name:\u003c\/p\u003e\n\n\u003cul\u003e\n\n\u003cli dir=\"ltr\"\u003e NVIDIA directly under \"NVIDIA DGX Spark\" (formerly Project Digits)\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003e ASUS as \"ASUS Ascent GX10\"\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003e HP as \"HP ZGX Nano AI Station G1n\"\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003e Dell as \"Dell Pro Max With GB10\"\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003e Lenovo as \"Lenovo ThinkStation PGX\"\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\n\u003cp\u003e In our opinion, these devices are largely comparable in terms of AI performance. We recommend customers purchase the brand of their choice and pay particular attention to other factors such as service provided by the respective brand.\u003c\/p\u003e\n\n \u003cp\u003eAs an ASUS IoT partner, we offer the ASUS Ascent GX10 variant of the NVIDIA DGX Spark in our shop. ASUS's strength lies in developing particularly optimized cooling concepts, which they have also implemented with the DGX Spark in the form of the ASUS Ascent GX10.\u003c\/p\u003e\n\n\u003ch2\u003e Software stack \/ supported software\u003c\/h2\u003e\n\n\u003cul\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA DGX OS\u003c\/strong\u003e (operating system based on Ubuntu)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA NIM\u003c\/strong\u003e (Microservices as building blocks for AI solutions)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA AI Blueprints\u003c\/strong\u003e (example applications, e.g. RAG workflows)\u003c\/li\u003e\n\n\u003cli\u003e \u003cspan\u003e\u003cstrong\u003ePyTorch\u003c\/strong\u003e (framework for machine learning and deep neural networks)\u003c\/span\u003e\n\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eTensorFlow\u003c\/strong\u003e ( \u003cspan\u003eFramework for Machine Learning and Deep Neural Networks\u003c\/span\u003e )\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA Riva\u003c\/strong\u003e (speech recognition, translation, and speech output)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA Holoscan\u003c\/strong\u003e (platform for sensor data processing)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA Metropolis\u003c\/strong\u003e (Vision AI Application Platform)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNVIDIA Isaac\u003c\/strong\u003e (platform for developing robots, e.g. autonomous vehicles)\u003c\/li\u003e\n\n\u003cli\u003e \n\u003cstrong\u003eCUDA\u003c\/strong\u003e (toolkit for accelerating and parallelizing computations on GPUs)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003ecuDNN\u003c\/strong\u003e (Acceleration of Deep Neural Networks on GPUs)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003ecuBLAS\u003c\/strong\u003e (Linear Algebra Frameworks for Accelerating AI and High Performance Computing (HPC) Applications)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eTensorRT\u003c\/strong\u003e (SDK for high-performance deep learning inferencing)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003eNCCL\u003c\/strong\u003e (NVIDIA Collective Communications library, for distributing computing power across multiple nodes)\u003c\/li\u003e\n\n\u003cli\u003e\n\n \u003cstrong\u003evLLM + ollama + llama.cpp\u003c\/strong\u003e - Large Language Model Inferencing Frameworks\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\n\u003cp\u003e\u003c\/p\u003e","brand":"ASUS","offers":[{"title":"1TB","offer_id":53110647849227,"sku":"nvidia-dgx-spark-1","price":4699.99,"currency_code":"EUR","in_stock":true},{"title":"4TB","offer_id":53110647881995,"sku":"nvidia-dgx-spark-2","price":6349.99,"currency_code":"EUR","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1560\/1473\/files\/ASUS_Ascent_GX10_with_NVIDIA-Logo.png?v=1758102014"},{"product_id":"nvidia-dgx-spark-4tb","title":"NVIDIA DGX Spark 4TB","description":"\u003ch1 dir=\"ltr\"\u003e\u003cspan\u003eNVIDIA DGX Spark 4 TB: AI Supercomputer\u003c\/span\u003e\u003c\/h1\u003e\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eThe NVIDIA DGX Spark is a compact AI supercomputer. A generous 128GB RAM configuration enables work with large LLMs with up to 200B parameters.\u003c\/span\u003e\u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eThe core of the NVIDIA DGX Spark is the \u003cspan class=\"fontstyle0\"\u003e\u003cstrong\u003eNVIDIA® GB10 Grace Blackwell Superchip\u003c\/strong\u003e, with 1 Petaflop of AI computing power (in FP4 precision).\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eThe NVIDIA DGX Spark platform is particularly suitable for AI developers and AI research. Thanks to its compact form factor, the NVIDIA DGX Spark fits on any desk.\u003c\/span\u003e\u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eAn essential part of the platform is the NVIDIA AI Software Stack, which enables a quick start with the best-supported platform in the LLM industry. Thanks to software compatibility with NVIDIA's larger platforms (e.g., H200 \/ B200), developed LLM solutions and applications (e.g., agentic AI) can be scaled directly and easily.\u003c\/span\u003e\u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eThanks to the built-in NVIDIA® ConnectX-7 network link, two NVIDIA DGX Spark systems can be connected to support even larger models with up to 405B parameters.\u003cbr\u003e\u003cbr\u003e\u003cspan style=\"color: rgb(0, 0, 0); background-color: rgb(255, 255, 0);\"\u003e\u003cstrong\u003eDelivery time\u003c\/strong\u003e\u003c\/span\u003e\u003cbr data-end=\"124\" data-start=\"121\"\u003e\u003cspan style=\"color: rgb(0, 0, 0); background-color: rgb(255, 255, 0);\"\u003eSmall quantities available \u003cstrong\u003edirectly\u003c\/strong\u003e from stock in Leipzig, Germany. The delivery time for larger quantities of the NVIDIA DGX Spark 4 TB is usually \u003cstrong\u003e4–6 weeks from receipt of order.\u003c\/strong\u003e Please consider this in your project planning.\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\u003ch2 dir=\"ltr\"\u003e\u003cspan\u003eTechnical Specifications\u003c\/span\u003e\u003c\/h2\u003e\n\u003cdiv align=\"left\" dir=\"ltr\"\u003e\n\n\u003ctable style=\"width: 100%; height: 790.334px;\"\u003e\n\n\u003ccolgroup\u003e \u003ccol style=\"width: 29.061%;\" width=\"139\"\u003e \u003ccol style=\"width: 70.6022%;\" width=\"421\"\u003e \u003c\/colgroup\u003e\n\n\u003ctbody\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cstrong\u003eFeature\u003c\/strong\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cstrong\u003eSpecification\u003c\/strong\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 55.1944px;\"\u003e\n\n\u003ctd style=\"height: 55.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eAI Computing Power (Tensor Performance)\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 55.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e1 Petaflops (FP4, sparse) (NVIDIA Grace Blackwell)\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eRAM\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e128GB RAM LPDDR5x (Unified memory)\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eRAM Bandwidth\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003e273 GB\/s (256-bit memory interface)\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 106.792px;\"\u003e\n\n\u003ctd style=\"height: 106.792px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eGPU\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 106.792px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003eNVIDIA Blackwell Architecture\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003e5th Generation Tensor Cores\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003e4th Generation RT Cores\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eCPU\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003e20 core ARM: 10 Cortex-X925 \u003c\/span\u003e\u003cspan dir=\"ltr\" role=\"presentation\"\u003e+ 10 Cortex-A725 ARM\u003c\/span\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eStorage\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x 4 TB NVME.M2 with self-encryption, Enterprise Grade Storage\u003cbr\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 55.1944px;\"\u003e\n\n\u003ctd style=\"height: 55.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eWLAN \u0026amp; Bluetooth\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 55.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eWi-Fi 7 + Bluetooth® 5.3 with Bluetooth LE\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 71.1944px;\"\u003e\n\n\u003ctd style=\"height: 71.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eNetwork\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 71.1944px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x 10 GBit LAN RJ45\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x ConnectX CX-7 Smart NIC (200 Gbps, 2 x QSFP)\u003cbr\u003e\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 110.389px;\"\u003e\n\n\u003ctd style=\"height: 110.389px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e \u003c\/p\u003e\n\n4x USB Type C, one port for power supply\u003cbr\u003e\n\u003cp dir=\"ltr\"\u003e \u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 106.792px;\"\u003e\n\n\u003ctd style=\"height: 106.792px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eAdditional Ports \u0026amp; Features\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 106.792px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x HDMI 2.1a, with HDMI multichannel Audio output\u003cbr\u003e(supports up to 8K display at 120 Hz and HDR)\u003cbr\u003e1 x NVENC, 1 x NVDEC\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e1 x Power button\u003c\/span\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eOperating System\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003eNVIDIA DGX™ OS (Ubuntu Linux)\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eDimensions\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e\u003cspan class=\"fontstyle0\"\u003e150 mm L x 150 mm W x 50.5 mm H\u003c\/span\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\u003ctr style=\"height: 35.5972px;\"\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003eWeight\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\u003ctd style=\"height: 35.5972px;\"\u003e\n\n\u003cp dir=\"ltr\"\u003e\u003cspan\u003e1.2 kg\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/td\u003e\n\n\n\u003c\/tr\u003e\n\n\n\u003c\/tbody\u003e\n\n\n\u003c\/table\u003e\n\n\n\u003c\/div\u003e\n\u003ch2 dir=\"ltr\"\u003e\u003cspan\u003eScope of Delivery\u003c\/span\u003e\u003c\/h2\u003e\n\u003cul\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003e\u003cspan\u003e\u003cstrong\u003e1\u003c\/strong\u003e × NVIDIA DGX Spark with 4 TB enterprise grade SSD\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003e\u003cspan\u003e\u003c\/span\u003ePower Supply\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003ePower Cable (EU) \u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\u003cspan\u003eNote: QSFP cable for the ConnectX CX-7 network interface, for connecting two NVIDIA DGX Spark systems, is not included - it is available separately. Compatible cables are:\u003c\/span\u003e\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003cul\u003e\n\n\u003cli style=\"list-style-type: none;\"\u003e\n\n\u003cul class=\"productTools-features gray-bullets\"\u003e\n\n\u003cli class=\"\"\u003e\u003cspan aria-expanded=\"true\"\u003eDGX Spark Connect Cable: NJAAK-N911\u003c\/span\u003e\u003c\/li\u003e\n\n\u003cli class=\"\"\u003e\u003cspan aria-expanded=\"false\"\u003eDGX Spark Connect Cable: NJAAK-0006\u003c\/span\u003e\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\n\n\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003cdiv style=\"text-align: start;\"\u003e\u003cimg style=\"margin-bottom: 16px; float: none; display: block; margin-left: auto; margin-right: auto;\" alt=\"\" src=\"https:\/\/cdn.shopify.com\/s\/files\/1\/1560\/1473\/files\/NVIDIA-DGX-SPARK-DOUBLE-Back_240x240.png?v=1762974315\"\u003e\u003c\/div\u003e\n\u003cp\u003e\u003cspan\u003e\u003cstrong\u003eNote\u003c\/strong\u003e: when ordering \u003cstrong\u003etwo\u003c\/strong\u003e NVIDIA DGX Spark 4 TB, we include a free QSFP cable during the promotion period.\u003c\/span\u003e\u003c\/p\u003e\n\u003ch2 dir=\"ltr\"\u003eAreas of Application\u003c\/h2\u003e\n\u003cul\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003e\u003cspan\u003eIndustry \u0026amp; Research\u003c\/span\u003e\u003cspan\u003e: Development and finetuning of AI LLM models with up to 200B parameters (or 405B when using two NVIDIA DGX Sparks)\u003c\/span\u003e\u003cspan\u003e\u003cbr\u003e\u003c\/span\u003e\u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003eUniversities \u0026amp; Colleges: Experiments and hands-on experience for students with large LLM models \u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\u003cli aria-level=\"1\" dir=\"ltr\"\u003e\n\n\u003cp role=\"presentation\" dir=\"ltr\"\u003eOffices, Business Environment: Running Agentic AI, especially for data protection - e.g., for processing customer inquiries, offline email processing, RAGs with sensitive company data. \u003c\/p\u003e\n\n\n\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003ch2 dir=\"ltr\"\u003e\u003cspan\u003eNVIDIA DGX Spark Product Variants\u003c\/span\u003e\u003c\/h2\u003e\n\u003cp dir=\"ltr\"\u003eThe NVIDIA DGX Spark platform is based on NVIDIA components at its core. Each variant features the Grace Blackwell Superchip GB10, as well as 128 GB RAM with 273 GB\/s memory bandwidth. Memory bandwidth has the greatest effect on token\/s throughput rates. \u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003eThe operating system (NVIDIA DGX OS, based on Ubuntu) and the NVIDIA AI software stack are the same in all cases.\u003c\/p\u003e\n\u003cp dir=\"ltr\"\u003eDifferent providers build a housing around it and market the device under their own brand name: \u003c\/p\u003e\n\u003cul\u003e\n\n\u003cli dir=\"ltr\"\u003eNVIDIA directly as \"NVIDIA DGX Spark\" (formerly Project Digits) \/ DGX Spark Founders Edition\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003eNVIDIA DGX Spark Founder's Edition\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003ePNY is in this case a reseller of NVIDIA, i.e. a PNY DGX Spark is an original NVIDIA DGX Spark\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003e\u003ca href=\"https:\/\/buyzero.de\/products\/asus-ascent-gx10-nvidia-dgx-spark\"\u003eASUS as \"ASUS Ascent GX10\"\u003c\/a\u003e\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003eHP as \"HP ZGX Nano AI Station G1n\"\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003eDell as \"Dell Pro Max With GB10\"\u003c\/li\u003e\n\n\u003cli dir=\"ltr\"\u003eLenovo as \"Lenovo ThinkStation PGX\"\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003cp\u003eIn our opinion, these devices are largely comparable in terms of AI performance. We recommend that customers buy the brand of their choice and pay particular attention to other factors such as service from the respective brand. \u003c\/p\u003e\n\u003cp\u003eAs an ASUS IoT partner, we also offer the \u003ca rel=\"noopener\" href=\"https:\/\/buyzero.de\/products\/asus-ascent-gx10-nvidia-dgx-spark\" target=\"_blank\"\u003eASUS Ascent GX 10 variant of the NVIDIA DGX Spark\u003c\/a\u003e in our shop. ASUS's strength is in developing optimized cooling concepts, which they have also implemented with the DGX Spark in the form of the \u003ca href=\"https:\/\/buyzero.de\/products\/asus-ascent-gx10-nvidia-dgx-spark\"\u003eASUS Ascent GX 10\u003c\/a\u003e. Another advantage of the ASUS Ascent GX10 is its lower price (4 TB variant of the ASUS Ascent GX 10 available on request, also more affordable).\u003c\/p\u003e\n\u003ch2\u003eSoftware Stack \/ Supported Software\u003c\/h2\u003e\n\u003cul\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA DGX OS\u003c\/strong\u003e (Operating system, based on Ubuntu)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA NIM\u003c\/strong\u003e (Microservices as building blocks for AI solutions)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA AI Blueprints\u003c\/strong\u003e (Example applications, e.g., RAG workflows)\u003c\/li\u003e\n\n\u003cli\u003e\u003cspan\u003e\u003cstrong\u003ePyTorch\u003c\/strong\u003e (Framework for Machine Learning and Deep Neural Networks)\u003c\/span\u003e\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eTensorFlow\u003c\/strong\u003e (\u003cspan\u003eFramework for Machine Learning and Deep Neural Networks\u003c\/span\u003e)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA Riva\u003c\/strong\u003e (Speech recognition, translation, and speech synthesis)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA Holoscan\u003c\/strong\u003e (Platform for sensor data processing)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA Metropolis\u003c\/strong\u003e (Vision AI application platform)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNVIDIA Isaac\u003c\/strong\u003e (Platform for developing robots, e.g., autonomous vehicles)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eCUDA\u003c\/strong\u003e (Toolkit for accelerating and parallelizing computations on GPUs)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003ecuDNN\u003c\/strong\u003e (Acceleration of Deep Neural Networks on GPUs)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003ecuBLAS\u003c\/strong\u003e (Linear Algebra Frameworks for accelerating AI and High Performance Computing (HPC) applications)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eTensorRT\u003c\/strong\u003e (SDK for high-performance deep learning inferencing)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003eNCCL\u003c\/strong\u003e (NVIDIA Collective Communications library, for distributing computing power across multiple nodes)\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cstrong\u003evLLM + ollama + llama.cpp\u003c\/strong\u003e - Large Language Model Inferencing Frameworks\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003ch2\u003eOther Product Designations\u003c\/h2\u003e\n\u003cul\u003e\n\n\u003cli\u003e\n\n\u003cspan class=\"fontstyle0\"\u003eDGXSPARK-FOUNEDIT-EU\u003c\/span\u003e \u003cbr\u003e\n\n\u003c\/li\u003e\n\n\u003cli\u003ePNY Part Number: NVDGXSPARK-PB \u003cspan class=\"fontstyle0\"\u003e\u003c\/span\u003e\n\n\u003c\/li\u003e\n\n\u003cli\u003e\n\n\u003cspan class=\"fontstyle0\"\u003ePNY NVIDIA DGX Spark GB10 128GB 1 Petaflop FP4 AI Compute \u003c\/span\u003e\u003cspan class=\"fontstyle0\"\u003e128 Gb memory 4TB enterprise grade storage CX7 networking EU \u003c\/span\u003e\u003cspan class=\"fontstyle0\"\u003ecable\u003c\/span\u003e \u003cspan class=\"fontstyle0\"\u003e\u003c\/span\u003e\n\n\u003c\/li\u003e\n\n\u003cli\u003e\u003cspan class=\"fontstyle0\"\u003eNVIDIA DGX Spark Founder's Edition\u003c\/span\u003e\u003c\/li\u003e\n\n\u003cli\u003e\u003cspan class=\"fontstyle0\"\u003e940-54242-0000\u003cbr\u003e\u003c\/span\u003e\u003c\/li\u003e\n\n\n\u003c\/ul\u003e\n\u003cp\u003e \u003c\/p\u003e","brand":"NVIDIA","offers":[{"title":"Default Title","offer_id":52681688678667,"sku":"nvidia-dgx-spark","price":6438.99,"currency_code":"EUR","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1560\/1473\/files\/NVIDIA-DGX-SPARK.png?v=1762974223"}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/1560\/1473\/collections\/NVIDIA-DGX-SPARK.png?v=1785513052","url":"https:\/\/buyzero.de\/en\/collections\/nvidia-dgx-spark.oembed","provider":"buyzero.de","version":"1.0","type":"link"}