Overview
The NVIDIA DGX Spark is a compact desktop AI supercomputer powered by the NVIDIA GB10 Grace Blackwell Superchip. Combining a 20-core Arm CPU, Blackwell GPU architecture, and 128GB unified LPDDR5x memory, it delivers up to 1 PFLOP of FP4 AI performance. Designed for local AI development, large language model inference, fine-tuning, autonomous agents, and advanced data science workloads.
Application & Use Cases:
- Local large language model development and inference
- AI model fine-tuning and prototyping
- Generative AI application development
- Autonomous AI agent development and deployment
- Computer vision and image processing workloads
- Robotics and edge AI application development
- Enterprise AI research and data science
- Multi-system AI clustering and distributed inference
- Multimodal AI development involving image, video, and language models
Features:
- NVIDIA Grace Blackwell architecture
- Integrated NVIDIA GB10 Grace Blackwell Superchip
- 20-core Arm CPU architecture
- Blackwell GPU with fifth-generation Tensor Cores
- Fourth-generation RT Cores
- Up to 1 PFLOP FP4 AI compute performance
- 128GB coherent unified LPDDR5x system memory
- 273 GB/s memory bandwidth
- Up to 4TB NVMe M.2 self-encrypting storage
- NVIDIA ConnectX-7 networking at up to 200 Gbps
- Wi-Fi 7 and Bluetooth 5.4 connectivity
- Four USB Type-C interfaces
- HDMI 2.1a display connectivity
- Support for up to four interconnected DGX Spark systems
- NVIDIA DGX OS and integrated AI software stack
Key Specifications:
- Product Name: NVIDIA DGX Spark
- Product Type: Desktop AI Supercomputer
- Product Architecture: NVIDIA Grace Blackwell
- System-on-Chip: NVIDIA GB10 Grace Blackwell Superchip
- GPU Architecture: NVIDIA Blackwell
- CPU Architecture: Arm
- CPU Core Count: 20 cores
- CPU Core Configuration: 10 Cortex-X925 + 10 Cortex-A725
- GPU: Integrated Blackwell GPU
- CUDA Cores: 6,144
- Tensor Cores: Fifth Generation
- RT Cores: Fourth Generation
- AI Compute Performance: Up to 1 PFLOP FP4 with sparsity
- System Memory: 128GB LPDDR5x
- Memory Architecture: Coherent unified system memory
- Memory Interface: 256-bit
- Memory Bandwidth: 273 GB/s
- Memory Channels: 16
- Memory Technology Speed: LPDDR5x-8533
- Storage Capacity: 4TB NVMe M.2
- Storage Security: Self-encrypting storage
- Networking Controller: NVIDIA ConnectX-7
- High-Speed Networking: Up to 200 Gbps
- Ethernet Interface: 1 x RJ-45 10 Gigabit Ethernet
- Ethernet Standard: 10 Gigabit Ethernet
- Wireless Networking: Wi-Fi 7
- Bluetooth: Bluetooth 5.4
- USB Ports: 4 x USB Type-C
- Display Interface: 1 x HDMI 2.1a
- Additional Display Output: Up to 3 x DisplayPort over USB-C (DP Alt Mode)
- Audio Output: HDMI multichannel audio
- Video Encoder: 1 x NVENC
- Video Decoder: 1 x NVDEC
- Operating System: NVIDIA DGX OS
- AI Framework Support: NVIDIA AI software stack, CUDA, PyTorch, and TensorRT-LLM
- Model Capacity: Up to 200 billion parameters on a single 128GB system, workload-dependent
- Multi-System Clustering: Up to 4 DGX Spark systems
- Chip TDP: 140 W
- Power Supply: 240 W external power supply
- Form Factor: Compact Desktop / Small Form Factor
- Dimensions (L x W x H): 150 x 150 x 50.5 mm
- System Weight: 1.2 kg
- Ideal Operating Temperature: 5°C to 30°C
- Operating Humidity: 10% to 90%, non-condensing