NVIDIA DGX Spark: Desktop AI Supercomputer Powered by Grace Blackwell
Compatible:
Cisco, Huawei, Arista, Juniper, HPE/Aruba, Dell, H3C, Ruijie, ZTE Nokia, Ericsson Extreme, Allied Telesis, NETGEAR, D-Link, TP-Link, MikroTik Finisar, Fortinet, Palo Alto Networks, Generic, Customized
Factory direct price
23 years of manufacturing expertise
100% detection
Each module is 100% tested
Technical Support
Professional free technical guidance
Spot stock
Always in stock, short delivery time
-
Product Description
Product Introduction
The NVIDIA DGX Spark is a compact desktop AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip, delivering up to 1 PetaFLOP of FP4 AI performance. Featuring 128GB of unified CPU-GPU memory, it eliminates traditional bottlenecks, enabling users to seamlessly prototype, fine-tune, and run massive AI models locally. Pre-installed with the NVIDIA AI software stack and DGX OS, it provides a plug-and-play environment for generative AI, physical AI, and autonomous agents, with seamless deployment to DGX Cloud.Application Scenarios
- AI Prototyping & Development: Rapidly build, test, and validate generative AI models and autonomous agents locally before cloud scaling.
- Large Model Fine-Tuning: Customize pre-trained models with up to 70 billion parameters for specific use cases.
- High-Performance Inference: Locally run models with up to 200B parameters, or scale to 405B parameters via dual-system interconnect.
- Advanced Data Science & Robotics: Accelerate data-intensive workloads, physical AI (e.g., NVIDIA GR00T N1), and edge computing simulations.
- Privacy-Sensitive Research: Securely process confidential data locally in fields like healthcare without relying on cloud infrastructure.
-
Product details
Feature Specification Core Superchip NVIDIA GB10 Grace Blackwell Superchip AI Compute Performance Up to 1 PetaFLOP (FP4 AI TOPS) GPU Architecture NVIDIA Blackwell GPU with 5th Gen Tensor Cores CPU Architecture NVIDIA Grace CPU (20-Core ARM: 10x Cortex-X925 + 10x Cortex-A725) Unified System Memory 128GB LPDDR5x (256-bit, 273 GB/s bandwidth) Storage Up to 4TB NVMe M.2 SSD (Self-Encrypting) Networking ConnectX-7 SmartNIC (200GbE), 10GbE RJ45, Wi-Fi 7, Bluetooth 5.3 I/O Ports 4x USB4 Type-C (40Gbps), 1x HDMI 2.1, 2x 200GbE QSFP Model Capacity Single: 200B Inference / 70B Fine-Tuning
Dual-Linked: 405B InferenceSoftware Stack NVIDIA DGX OS, PyTorch, NeMo, RAPIDS, NVIDIA NIM Dimensions & Weight 150mm x 150mm x 50.5mm / 1.2 kg Power Consumption ~170W TDP -
Applications
-
Compitibility Test
Related Products
Quality Certification
Please contact us to download the certification certificate.
Please Note: Installing a third party transceiver does not void your network equipment warranty. Network equipment manufacturers all have guidelines stating that warranty supportontheirproductswillnotbeaffected.
Questions & Answers
Q20: How to contact the brand for order-related issues?
A20: For order inquiries and status updates, please contact our official customer service directly with your order information for quick assistance.
Q19: What exclusive services are provided to cooperative customers?
A19: Cooperative customers enjoy full-range services including dedicated consultation, technical guidance, after-sales support and real-time information updates.
Q18: How can I get support for product usage problems?
A18: For product usage and technical questions, you can get professional support through our official hotline, online consultation or message system.
Q17: What are the quality guarantee and after-sales policy?
A17: We implement strict quality control and provide official quality assurance. We follow a unified after-sales policy to offer customers reliable and complete support.
Q16: Does the brand support OEM/ODM customization?
A16: With a mature operating system of 23 years, we support OEM/ODM cooperation and provide customized production and solutions based on customer requirements.
Online Message















