{"product_id":"pny-nvidia-dgx-spark-gb10-128gb-ai-supercomputer","title":"PNY Nvidia DGX Spark GB10 128GB Ai Supercomputer","description":"\u003cp\u003e\u003cstrong\u003eDesktop AI supercomputer | 1 petaFLOP FP4 performance | 128 GB unified memory | NVIDIA Grace Blackwell architecture | Prototype, fine-tune and deploy large AI models locally\u003c\/strong\u003e\u003c\/p\u003e\n\u003ch3\u003eQuick Specs\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003eAchieve up to 1 petaFLOP of AI compute performance at FP4 precision with the NVIDIA Grace Blackwell architecture — data-centre power on your desk.\u003c\/li\u003e\n\u003cli\u003eRun large AI models of up to 200 billion parameters locally, backed by 128 GB of unified LPDDR5x system memory and 273 GB\/s memory bandwidth.\u003c\/li\u003e\n\u003cli\u003eStore massive datasets and model checkpoints on a 4 TB self-encrypting NVMe M.2 drive — fast, capacious, and secure.\u003c\/li\u003e\n\u003cli\u003eConnect two DGX Spark units via the ConnectX-7 Smart NIC to tackle AI models of up to 405 billion parameters in a collaborative desktop cluster.\u003c\/li\u003e\n\u003cli\u003eWork wirelessly and freely with Wi-Fi 7 and Bluetooth 5.3, while the compact 15 × 15 × 5.05 cm form factor keeps your workspace uncluttered.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3\u003eA new class of computer for AI builders and researchers\u003c\/h3\u003e\n\u003cp\u003eThe PNY NVIDIA DGX Spark GB10 redefines what a desktop workstation can do. Designed for developers, data scientists, and AI researchers who need real supercomputer capability without a server room, it delivers up to 1 petaFLOP of AI compute at FP4 precision from a compact, champagne-gold enclosure that weighs just 1.2 kg. Powered by the NVIDIA Grace Blackwell architecture — the same generation driving the world's most advanced data centres — the DGX Spark lets you prototype, fine-tune, and run inference on large language models and generative AI workloads entirely on your desktop, then deploy seamlessly to the data centre or cloud when you are ready to scale.\u003c\/p\u003e\n\u003ch3\u003e128 GB of unified memory: no model too large for your desk\u003c\/h3\u003e\n\u003cp\u003eTired of hitting memory walls mid-training and waiting on cloud queues to continue your work? The DGX Spark's 128 GB of LPDDR5x unified system memory — shared coherently between the CPU and GPU — removes that bottleneck entirely. With a 256-bit memory interface delivering 273 GB\/s of bandwidth, you can run AI models of up to 200 billion parameters locally in a single system. Link two DGX Spark units together via NVIDIA ConnectX networking and the combined capacity extends to AI models of up to 405 billion parameters, giving your team a compact, high-performance AI cluster without the infrastructure overhead.\u003c\/p\u003e\n\u003ch3\u003eGrace Blackwell silicon: 5th-generation Tensor Cores at your fingertips\u003c\/h3\u003e\n\u003cp\u003eAt the heart of the DGX Spark is the NVIDIA GB10 Superchip, combining a 20-core Arm CPU — ten Cortex-X925 performance cores paired with ten Cortex-A725 efficiency cores — with a Blackwell-generation GPU featuring 5th-generation Tensor Cores and 4th-generation RT Cores. This architecture is purpose-built for the compute demands of modern AI: transformer training, large-scale inference, retrieval-augmented generation, and multi-modal workloads all benefit from the raw arithmetic throughput that previously required a rack of servers. The preinstalled NVIDIA AI software stack — including frameworks, libraries, and pretrained models — means you can be productive from the moment you power on.\u003c\/p\u003e\n\u003ch3\u003eEnterprise-grade connectivity in a mini form factor\u003c\/h3\u003e\n\u003cp\u003eDespite its compact footprint, the DGX Spark is exceptionally well connected. Four USB Type-C ports (three available for peripherals, one powering the unit), a single HDMI 2.1a output supporting up to 8K displays at 120 Hz with HDR, a 10 GbE RJ-45 port for reliable wired networking, and a ConnectX-7 Smart NIC capable of up to 100 GbE ensure that data moves as fast as your models demand. Wi-Fi 7 and Bluetooth 5.3 keep the unit cable-light when you need flexibility. The 4 TB self-encrypting NVMe M.2 drive safeguards your proprietary models and datasets at rest, meeting the security expectations of professional AI development environments.\u003c\/p\u003e\n\u003ch3\u003eFeatures\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003eNVIDIA Grace Blackwell architecture delivers up to 1 petaFLOP of FP4 AI compute performance from a desktop-sized unit.\u003c\/li\u003e\n\u003cli\u003e128 GB of unified LPDDR5x system memory with 273 GB\/s bandwidth supports AI models of up to 200 billion parameters in a single system.\u003c\/li\u003e\n\u003cli\u003e5th-generation Tensor Cores and 4th-generation RT Cores accelerate transformer training, inference, and generative AI workloads.\u003c\/li\u003e\n\u003cli\u003eConnectX-7 Smart NIC enables dual-unit networking for AI models of up to 405 billion parameters across two connected DGX Spark systems.\u003c\/li\u003e\n\u003cli\u003e4 TB self-encrypting NVMe M.2 storage keeps proprietary model data and datasets secure and immediately accessible.\u003c\/li\u003e\n\u003cli\u003eFour USB Type-C ports, HDMI 2.1a (8K\/120 Hz HDR), 10 GbE RJ-45, Wi-Fi 7, and Bluetooth 5.3 for comprehensive desktop connectivity.\u003c\/li\u003e\n\u003cli\u003ePreinstalled DGX OS with the full NVIDIA AI software stack — frameworks, libraries, and pretrained models included out of the box.\u003c\/li\u003e\n\u003cli\u003eCompact mini form factor (15 × 15 × 5.05 cm, 1.2 kg) in a distinctive champagne-gold finish.\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3\u003eSpecifications\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003eArchitecture: NVIDIA Grace Blackwell\u003c\/li\u003e\n\u003cli\u003eGPU: Blackwell Architecture\u003c\/li\u003e\n\u003cli\u003eCPU: 20 core Arm, 10 Cortex-X925 + 10 Cortex-A725\u003c\/li\u003e\n\u003cli\u003eCUDA Cores: Blackwell Generation\u003c\/li\u003e\n\u003cli\u003eTensor Cores: 5th Generation\u003c\/li\u003e\n\u003cli\u003eRT Cores: 4th Generation\u003c\/li\u003e\n\u003cli\u003eTensor Performance: 1 PFLOP\u003c\/li\u003e\n\u003cli\u003eSystem Memory: 128GB LPDDR5x, unified system memory\u003c\/li\u003e\n\u003cli\u003eMemory Interface: 256-bit\u003c\/li\u003e\n\u003cli\u003eMemory Bandwidth: 273 GB\/s\u003c\/li\u003e\n\u003cli\u003eOS: DGX OS\u003c\/li\u003e\n\u003cli\u003eStorage: 4 TB NVME.M2 with self-encryption\u003c\/li\u003e\n\u003cli\u003eUSB: 4x USB Type-C\u003c\/li\u003e\n\u003cli\u003eEthernet: 10 GbE 1x RJ-45 connector\u003c\/li\u003e\n\u003cli\u003eNIC: ConnectX-7 Smart NIC\u003c\/li\u003e\n\u003cli\u003eWi-Fi: Wi-Fi 7\u003c\/li\u003e\n\u003cli\u003eBluetooth: BT 5.3\u003c\/li\u003e\n\u003cli\u003eAudio output: HDMI® multichannel audio output\u003c\/li\u003e\n\u003cli\u003ePower Consumption: TBD\u003c\/li\u003e\n\u003cli\u003eDisplay Connectors: 1x HDMI®™ 2.1a\u003c\/li\u003e\n\u003cli\u003eNVENC | NVDEC: 1x | 1x\u003c\/li\u003e\n\u003cli\u003eProduct Dimensions: 15 x 15 x 5 cm\u003c\/li\u003e\n\u003cli\u003eProduct Weight: 1200g\u003c\/li\u003e\n\u003c\/ul\u003e\n\u003ch3\u003eWhat's in the Box\u003c\/h3\u003e\n\u003cul\u003e\n\u003cli\u003eNVIDIA DGX Spark – Gold\/Champagne x1\u003c\/li\u003e\n\u003cli\u003ePower Supply x1\u003c\/li\u003e\n\u003cli\u003eQuick Start Card x1\u003c\/li\u003e\n\u003cli\u003eSupport Card x1\u003c\/li\u003e\n\u003c\/ul\u003e","brand":"PNY","offers":[{"title":"Default Title","offer_id":49082155237476,"sku":"DGXSPARK-FOUNEDIT","price":103980.0,"currency_code":"ZAR","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0762\/6230\/9988\/files\/DGXSPARK-FOUNEDIT_PNY_Nvidia_DGX_Spark_GB10_128GB_Ai_Super_wr_01.png?v=1787843772","url":"https:\/\/www.primelaptopparts.co.za\/products\/pny-nvidia-dgx-spark-gb10-128gb-ai-supercomputer","provider":"Prime Laptop Parts","version":"1.0","type":"link"}