NVIDIA DGX Spark
Price range: $2,999.00 through $4,999.00
We are an NVIDIA-approved partner, authorized to sell a limited quantity of NVIDIA DGX Spark units for select early adopters.
NVIDIA DGX Spark offers an economical experimentation ground for prototype models, freeing up valuable compute resources in cluster environments. Its compatibility with the NVIDIA AI platform software architecture allows for seamless migration of models from your desktop to DGX Cloud or any accelerated cloud or data center infrastructure with virtually no code changes.
Key Features:
- Grace Blackwell AI Superchip: Powered by the NVIDIA GB10 Grace Blackwell Superchip, delivering 1 petaFLOP of AI performance at FP4 precision.
- High-Performance Processing: Features a powerful NVIDIA Blackwell GPU with fifth-generation Tensor Cores and a 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725) to supercharge data preprocessing, orchestration, and real-time inferencing.
- Unified System Memory: Equipped with 128 GB of coherent LPDDR5x unified system memory, enabling the development and testing of AI models up to 200 billion parameters directly from your desktop.
- Fast Memory Bandwidth: Offers 273 GB/s memory bandwidth.
- Flexible Storage: Available with 1 or 4 TB NVMe M.2 storage with self-encryption.
- Advanced Connectivity: Includes 4x USB Type-C ports, 1x RJ-45 10 GbE Ethernet, NVIDIA ConnectX-7 Smart NIC for up to 200GbE clustering, WiFi 7, and Bluetooth 5.3.
- Compact Design: A power-efficient, desk-friendly form factor (150mm L x 150mm W x 50.5mm H, 1.2 kg).
- NVIDIA AI Software Stack: Comes pre-installed with NVIDIA’s enterprise-grade software stack, including NVIDIA DGX OS, optimized for AI workloads and supporting frameworks like TensorFlow, PyTorch, MATLAB, and more.
Technical Specifications:
| Architecture | NVIDIA Grace Blackwell |
|---|---|
| GPU | NVIDIA Blackwell Architecture |
| CPU | 20-core Arm (10 Cortex-X925 + 10 Cortex-A725) |
| Tensor Performance | 1000 AI TOPS / 1 petaFLOP (FP4) |
| System Memory | 128 GB LPDDR5x, Coherent Unified System Memory |
| Memory Bandwidth | 273 GB/s |
| Storage | 1 or 4 TB NVMe M.2 with self-encryption |
| Networking | 1x 10 GbE, ConnectX-7 Smart NIC, WiFi 7, Bluetooth 5.3 |
| USB Ports | 4x USB Type-C |
| Display Connectors | 1x HDMI 2.1a |
| Dimensions (L x W x H) | 150 mm x 150 mm x 50.5 mm |
| Weight | 1.2 kg |
| Operating System | NVIDIA DGX™ OS |
Accelerate All AI Workloads:
- Prototyping: Develop, test, and validate AI models and applications with ease.
- Fine-Tuning: Improve pre-trained models by fine-tuning models up to 70 billion parameters.
- Inference: Test, validate, and inference AI models up to 200 billion parameters.
- Data Science: Accelerate end-to-end data science workflows with NVIDIA RAPIDS™.
- Edge Applications: Develop robotics, smart cities, and computer vision solutions using NVIDIA AI frameworks like Isaac™ and Metropolis.
| NVIDIA Founders Edition |
NVIDIA DGX Spark – 4TB ,NVIDIA DGX Spark Bundle 2 NVIDIA DGX Spark Units – 4TB with Connecting Cable |
|---|
4 reviews for NVIDIA DGX Spark
Related products
Bosgame M5 AI Mini Desktop Ryzen AI Max+ 395
EVO-X2 AI Mini PC – AMD Ryzen™ AI Max+ 395
The EVO-X2 redefines what's possible for AI-enabled mini PCs, supporting local inference of large AI models. With its powerful heterogeneous computing architecture, it delivers exceptional AI performance, outperforming the NVIDIA GeForce RTX 4090 by 2.2x in LM Studio benchmarks while consuming significantly less power. Experience fast, seamless performance for natural language processing, text generation, intelligent Q&A, image recognition, and AI-driven artwork.
Emily Chang –
The performance is unmatched. It’s a game-changer for AI startups.
James Wilson –
Seamless AI integration and unmatched power. NVIDIA has outdone themselves!
David Park –
Finally, enterprise-grade AI computing at a reasonable price point.
Dr. Sarah Chen –
Incredible performance for AI prototyping—nothing else compares!