LEADTEK NVIDIA DGX Spark DGX Mini PC AI (CO2-003735)

​

Compact AI Supercomputer Powered by NVIDIA Grace Blackwell for Generative AI and LLM Development

The LEADTEK NVIDIA DGX Spark (CO2-003735) is a compact AI supercomputer designed for AI developers, researchers, data scientists, and enterprises working with Generative AI, Large Language Models (LLMs), machine learning, and advanced AI workloads. Built on the NVIDIA Grace Blackwell architecture, DGX Spark brings data-center-class AI computing directly to the desktop.

Powered by the NVIDIA GB10 Grace Blackwell Superchip, 128GB LPDDR5X unified memory, and up to 4TB NVMe SSD storage, the system delivers up to 1,000 TOPS of AI performance and supports AI models with up to 200 billion parameters locally. Two DGX Spark systems can be linked together to support models up to 405 billion parameters.

Key Features of LEADTEK NVIDIA DGX Spark

​

NVIDIA GB10 Grace Blackwell Superchip

Powered by the NVIDIA GB10 Superchip, integrating NVIDIA Grace CPU and NVIDIA Blackwell GPU into a single AI computing platform.

Up to 1,000 TOPS AI Performance

Delivers up to 1,000 TOPS of FP4 AI compute performance, making it ideal for Generative AI, LLM inference, AI agents, deep learning, and computer vision applications.

NVIDIA Blackwell GPU

Features NVIDIA Blackwell architecture with fifth-generation Tensor Cores for accelerated AI training and inference workloads.

20-Core NVIDIA Grace CPU

Includes a 20-core Arm-based NVIDIA Grace CPU optimized for AI development, model deployment, and large-scale data processing.

128GB Unified LPDDR5X Memory

Equipped with 128GB of coherent unified memory, enabling seamless CPU-GPU memory sharing and efficient handling of large AI models.

4TB NVMe SSD Storage

Provides up to 4TB of high-speed NVMe storage for AI datasets, model repositories, and enterprise-scale workloads.

Supports Up to 200B-Parameter Models

Capable of running AI models up to 200 billion parameters locally and up to 405 billion parameters when two DGX Spark systems are connected via NVIDIA ConnectX networking.

High-Speed Networking

Features NVIDIA ConnectX-7 SmartNIC, 10Gb Ethernet, and Wi-Fi 7 connectivity for AI clusters and high-performance data transfer.

NVIDIA AI Software Ecosystem

Ships with NVIDIA DGX OS and supports CUDA, TensorRT, NIM, NeMo, PyTorch, TensorFlow, and other leading AI frameworks.

Ideal Applications

​

• Large Language Models (LLMs)
• Generative AI
• AI Agent Development
• Machine Learning
• Deep Learning
• Data Science
• Computer Vision
• Robotics
• Smart Cities
• Edge AI
• AI Model Fine-Tuning
• Research and Development

Why Choose LEADTEK NVIDIA DGX Spark

​

The LEADTEK NVIDIA DGX Spark (CO2-003735) combines NVIDIA Grace Blackwell architecture, 1,000 TOPS AI performance, 128GB unified memory, and enterprise AI software support into a compact desktop platform. It is an ideal solution for developers, researchers, and organizations seeking powerful local AI computing for model development, inference, fine-tuning, and next-generation AI innovation.