The NVIDIA DGX™ Spark with the powerful NVIDIA GB10 Grace Blackwell superchip delivers outstanding AI performance of 1,000 AI TOPS - in a compact and energy-efficient form factor. With 128 GB of memory and the pre-installed NVIDIA AI software package, the system supports local development, fine-tuning and inference of state-of-the-art AI models with up to 200 billion parameters - including models from DeepSeek, Meta, Google and other leading providers. It can be deployed flexibly in the data centre or in the cloud.
As an NVIDIA partner, we are your contact for powerful and space-saving NVIDIA DGX™ Spark solutions for your company.
Experience AI performance of up to 1 PetaFLOPS at FP4 precision with the Grace Blackwell architecture.
Run AI development and AI testing workloads with AI models with up to 200 billion parameters on your desktop with large, unified system memory.
High-performance NVIDIA ConnectX networks enable the connection of two NVIDIA DGX Spark systems to work with AI models with up to 405 billion parameters.
Use a full-stack solution for generative AI workloads, including tools, frameworks, libraries and pre-trained models.
| Feature | NVIDIA DGX™ Spark |
|---|---|
| GPU | Blackwell architecture |
| CPU | 20 compute units Arm, 10 Cortex-X925 + 10 Cortex-A725 |
| CUDA computing units | Blackwell generation |
| Tensor computing units | 5. Generation |
| RT computing units | 4. Generation |
| Tensor performance | 1.000 AI TOPS |
| Memory | 128 GB LPDDR5x, unified system memory |
| Memory interface | 256 bit |
| Memory bandwidth | 273 GB/s |
| Data storage | 1 or 4 TB NVMe.M2 with self-encryption |
| USB | 4 x USB4 Type-C (up to 40 GB/s) |
| Ethernet | 1 x RJ45 connection 10 GbE |
| NIC | ConnectX-7 Smart NIC |
| WLAN | WiFi 7 |
| Bluetooth | BT 5.3 |
| Power consumption | 170 watts |
| Screen connections | 1x HDMI 2.1a |
| NVENC | NVDEC | 1x | 1x |
| Operating system | NVIDIA DGX™ Base OS, Ubuntu Linux |
| Dimensions | 150 mm x 150 mm x 50.5 mm (LxWxH) |
| Weight | 1.2 kg |
