NVIDIA DGX Spark AI Supercomputer 4TB | 128GB Memory

• Generative AI development — Supports local development and testing of generative AI applications.
• Large language model inference — Provides an AI computing platform for local LLM inference workflows.
• AI model fine-tuning — Supports model adaptation and experimentation within the system's available resources.
• Deep learning research — Provides CPU, GPU and unified memory resources for neural-network development.
• Data science projects — Offers local storage and computing for data-analysis and AI pipelines.
• AI prototyping labs — Supports early-stage development before suitable workloads are moved to larger environments.
• Computer vision development — Supports computer-vision pipeline development and testing.
• Robotics development — Provides a desktop platform for supported robotics AI development workflows.
• Diffusion model experimentation — Supports development and testing of diffusion-model workloads.
• Enterprise AI evaluation — Allows technical teams to evaluate AI applications in a local environment.
• Research and development teams — Offers a compact system for AI experimentation and technical investigation.
• AI dataset preparation — Provides 4TB of local SSD capacity for datasets and related project files.
• Model checkpoint storage — Provides local storage for supported model files and checkpoints.
• Cloud-to-local development workflows — Supports local prototyping before scaling suitable workloads to cloud environments.
• Data-center AI planning — Can be used to prototype workloads intended for subsequent evaluation on larger DGX infrastructure.

• Grace-Blackwell architecture - Combines Arm CPU and Blackwell GPU computing in one system.
• GB10 Superchip - Integrates CPU and GPU resources within a compact desktop platform.
• Up to 1 PFLOP FP4 performance - Supports demanding AI development and inference workloads.
• 20-core Arm CPU - Provides 10 Cortex-X925 and 10 Cortex-A725 CPU cores for processing tasks.
• 128GB unified LPDDR5x memory - Provides shared CPU-GPU memory for supported AI workloads.
• 273GB/s memory bandwidth - Specifies the available memory bandwidth for data-intensive processing.
• 4TB NVMe M.2 SSD - Offers local storage for datasets, models and checkpoints.
• Self-encrypting SSD - Provides storage encryption capability as identified in the supplied specifications.
• NVIDIA DGX OS - Supplies the operating system environment for AI development workflows.
• ConnectX-7 networking - Provides a Smart NIC specified for connectivity up to 200Gbps.
• 10GbE RJ-45 interface - Supports wired Ethernet connectivity through a standard RJ-45 port.
• Wi-Fi 7 and Bluetooth 5.4 - Adds wireless networking and Bluetooth connectivity options.
• Four USB-C ports - Provides USB-C connections for supported peripherals and devices.
• HDMI 2.1a output - Provides a display interface for a compatible monitor.

• Brand: NVIDIA
• Product Name: NVIDIA DGX Spark AI Supercomputer
• Product Series: DGX Spark AI
• Model / Part Number: 940-54242-0007-000
• EAN / UPC: 810152850381
• Product Type: Compact desktop AI supercomputer
• Architecture: NVIDIA Grace-Blackwell
• Superchip: NVIDIA GB10
• CPU: 20-core Arm CPU
• CPU Core Configuration: 10 × Cortex-X925 + 10 × Cortex-A725
• GPU Architecture: NVIDIA Blackwell
• Tensor Cores: 5th Generation
• RT Cores: 4th Generation
• AI Performance: Up to 1 PFLOP FP4
• System Memory: 128GB unified LPDDR5x
• Memory Bandwidth: 273GB/s
• Storage Capacity: 4TB
• Storage Type: NVMe M.2 self-encrypting SSD
• Wired Networking: 10GbE RJ-45
• Additional Networking: ConnectX-7 Smart NIC, up to 200Gbps
• Wireless Connectivity: Wi-Fi 7 and Bluetooth 5.4
• USB Connectivity: 4 × USB-C ports
• Display Interface: HDMI 2.1a
• Operating System: NVIDIA DGX OS
• Power Consumption: 240W
• Dimensions: 150 × 150 × 50.5mm

WHY CHOOSE THIS PRODUCT?

• Local AI development - Supports teams that want to develop and test AI workloads on a desktop system.
• Unified memory capacity - Provides 128GB of shared CPU-GPU memory for supported processing tasks.
• Substantial local storage - The 4TB SSD accommodates datasets, models and development checkpoints.
• Compact deployment - Its stated dimensions support placement in desktop labs and development workspaces.
• AI-focused computing architecture - Combines Grace CPU and Blackwell GPU technologies in the GB10 Superchip.
• Defined AI performance specification - Up to 1 PFLOP FP4 performance helps buyers evaluate AI computing requirements.
• Multiple networking options - Includes 10GbE Ethernet and a ConnectX-7 Smart NIC specified for up to 200Gbps.
• Integrated system software - NVIDIA DGX OS provides the operating system for the intended AI workflows.
• Flexible development applications - Supports work involving generative AI, LLMs, data science and computer vision.
• One-year warranty - Provides a stated warranty period for procurement planning.