NVIDIA DGX OS System
Work seamlessly. Includes a custom Linux distribution optimized exclusively for AI workloads, with pre-configured drivers and direct access to the NVIDIA AI Enterprise platform.
$0.00
Your cart is empty
Join our community for exclusive offers and updates.
No, thank you. I'll pay full price.
OURBIZ-US Shop
Storage: 128GB / 1TB
The NVIDIA DGX Spark is the ultimate desktop personal supercomputer, designed for developers, researchers, and data scientists to build and run AI models at maximum speed locally. At the heart of this unit lies the Grace Blackwell GB10 Superchip, which combines a powerful 20-core ARM processor with the advanced Blackwell GPU architecture, offering up to 1 petaFLOP of FP4 performance to accelerate your workflows in an ultra-compact form factor.
This exclusive Gold-colored unit is ready for the highest corporate and research demands. With 128 GB of LPDDR5x unified memory coherent between the CPU and GPU, it effortlessly handles massive language models and large data batches. In addition, it features immense 4 TB NVMe M.2 storage, ultra-fast Wi-Fi 7 connectivity, and 10 GbE networking, integrating the NVIDIA DGX OS ecosystem out of the box for immediate execution.
The GB10 Superchip includes ARM processors (Cortex-X925 and Cortex-A725) along with 5th generation Tensor Cores. It achieves up to 1 PetaFLOP (FP4) of pure power, optimizing critical tasks for prototyping, fine-tuning, and model inference.
Unlike fragmented memory in traditional PCs, the DGX Spark uses 128 GB of LPDDR5x shared memory coherently between the CPU and GPU (273 GB/s bandwidth). It is the gold standard for handling Large Language Models (LLMs) locally.
Store huge training datasets and complex applications with ultra-fast read and write speeds, thanks to its 4 TB NVMe M.2 Solid State Drive (SSD) with native encryption.
Scale your projects in the cloud or local clusters with ease. It integrates the NVIDIA ConnectX-7 smart adapter, 10 GbE RJ-45 Ethernet port, latest-generation Wi-Fi 7, and Bluetooth 5.4.
The NVIDIA DGX Spark redefines what's possible directly from your desktop, bringing data center tools into a robust, ready-to-use ecosystem.
Work seamlessly. Includes a custom Linux distribution optimized exclusively for AI workloads, with pre-configured drivers and direct access to the NVIDIA AI Enterprise platform.
The entire Blackwell architecture condensed into a desktop unit. Ideal for data scientists who require high local performance for research, testing, and production tasks.
Connect high-demand monitors and peripherals via its four USB Type-C ports (with power delivery support) and its HDMI 2.1a video output.
| Model | NVIDIA DGX Spark (Gold Edition / SKU: 6676779) |
|---|---|
| Processor (CPU) | 20-core ARM Architecture (10 Cortex-X925 + 10 Cortex-A725) |
| Graphics (GPU) | NVIDIA Blackwell Architecture with 5th Generation Tensor Cores and 4th Generation RT Cores |
| AI Performance | Up to 1 PetaFLOP (FP4) |
| RAM Memory | 128 GB LPDDR5x Unified System Memory (273 GB/s bandwidth) |
| Storage | 4 TB NVMe M.2 SSD |
| Networking & I/O | 1x RJ-45 (10 GbE), NVIDIA ConnectX-7 Smart NIC, Wi-Fi 7, Bluetooth 5.4, 4x USB Type-C, 1x HDMI 2.1a |
| Operating System | NVIDIA DGX OS (Linux-based, optimized for AI) |
| Dimensions | 150 x 150 x 50.5 mm |
| Weight | 1.2 kg |
| Power | 240W external power supply required (included) |
Unlike traditional computers that separate system memory (RAM) from video card memory (VRAM), the DGX Spark shares its 128 GB of memory coherently between the ARM processor and the GPU. This allows huge Artificial Intelligence models to be loaded without delays when moving data between components.
This product includes one (1) NVIDIA DGX Spark supercomputer unit in Gold finish, configured with 128 GB of unified RAM, 4 TB of NVMe SSD storage, and its respective original 240W power supply.
It is not designed for gaming or traditional office use. Although it has exceptionally powerful hardware, the Grace Blackwell architecture and its native environment (NVIDIA DGX OS) are meticulously structured for data science, Large Language Model (LLM) development, and local neural network prototyping.