NVidia DGX Spark - Personal AI Supercomputer 4TB
NVidia DGX Spark - Personal AI Supercomputer 4TB
If you have any questions, you are always welcome to contact us. We'll get back to you as soon as possible, within 24 hours on weekdays.
-
Shipping Information
Use this text to answer questions in as much detail as possible for your customers.
-
Customer Support
Use this text to answer questions in as much detail as possible for your customers.
-
FAQ’s
Use this text to answer questions in as much detail as possible for your customers.
-
Contact Us
Use this text to answer questions in as much detail as possible for your customers.
The NVidia DGX Spark is a personal AI supercomputer designed to empower developers and organizations to build and run AI models with unprecedented efficiency. This DGX personal AI computer offers an economical experimentation ground, allowing for local development and prototyping of models before seamless deployment anywhere at scale.
Key Features at a Glance
- Designed as a personal AI computer for building and running AI.
- Features 128 GB of unified system memory.
- Supports FP4 data format for AI models up to 200B parameters.
- Includes built-in NVIDIA ConnectX network technology.
- Leverages the NVIDIA AI platform software architecture.
- Enables local prototyping, fine-tuning, and inferencing of large models.
Economical Prototyping for AI Models
The NVIDIA DGX Spark serves as an economical and powerful experimentation platform for developing prototype AI models. This personal AI computer frees up valuable compute resources in larger cluster environments, which are better suited for the training and deployment of production-grade models.
Seamless Workflow from Desktop to Cloud
Built on the NVIDIA AI platform software architecture, the DGX Spark allows users to effortlessly transfer their work from a desktop environment to DGX Cloud or other accelerated cloud and data center infrastructures. This integration simplifies the process of prototyping, fine-tuning, and iterating on AI models at any scale.
Local Development of Large AI Models
Designed to address the demands of modern AI, the NVidia DGX Spark enables developers to work with substantial AI models directly from their desktop. Featuring 128 GB of unified system memory and FP4 data format support, it can handle models up to 200B parameters for prototyping, fine-tuning, and inferencing. Built-in NVIDIA ConnectX network technology also allows two systems to be connected for working with even larger models like Llama 3.1 405B.
Technical Specifications
| Architecture | NVIDIA Grace Blackwell |
| GPU | NVIDIA Blackwell Architecture |
| CPU |
20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm |
| CUDA Cores | NVIDIA Blackwell Generation |
| Tensor Cores | 5th Generation |
| RT Cores | 4th Generation |
| Tensor Performance | 1 PFLOP |
| System Memory | 128 GB LPDDR5x, coherent unified system memory |
| Memory Interface | 256-bit |
| Memory Bandwidth | Up to 273 GB/s |
| Storage | 4 TB NVME.M2 with self-encryption |
| USB | 4x USB TypeC |
| Ethernet | 1x RJ-45 connector 10 GbE |
| NIC | ConnectX-7 NIC @ 200 Gbps |
| Wi-Fi | WiFi 7 |
| Bluetooth | BT 5.4 w/LE |
| Audio-output | HDMI multichannel audio output |
| Power Supply | 240 W |
| GB10 TDP* | 140 W |
| Display Connectors | 1x HDMI 2.1a |
| NVENC | NVDEC | 1x | 1x |
| OS | NVIDIA DGX™ OS |
| System Dimensions | 150 mm L x 150 mm W x 50.5 mm H |
| System Weight | 1.2 kg |
Frequently Asked Questions
Q: What is the primary purpose of the NVIDIA DGX Spark?
A: It is a personal AI computer designed for organizations and developers to build, run, prototype, fine-tune, and inference AI models.
Q: What is the memory capacity and model parameter support of the DGX Spark?
A: The DGX Spark features 128 GB of unified system memory and supports AI models up to 200B parameters, utilizing the FP4 data format.
Q: How does the DGX Spark facilitate working with even larger AI models?
A: With built-in NVIDIA ConnectX network technology, two DGX Spark systems can be connected to work on larger models, such as Llama 3.1 405B.
