MikhbarMIKHBAR
Artificial Intelligence

NVIDIA DGX Spark 64GB Launches Local AI Development

The new 64GB configuration of NVIDIA DGX Spark brings on-device AI agent capabilities, clustering features, and the full NVIDIA AI software stack to developers starting this October.

NVIDIA DGX Spark 64GB Launches Local AI Development

Introduction to Personal AI Supercomputing

Local artificial intelligence is expanding as open models shrink to fit personal devices, allowing developers to run workloads directly on their hardware. To meet this demand, NVIDIA has announced the NVIDIA DGX Spark 64GB, a new personal AI supercomputer configuration designed to give developers, researchers, and AI enthusiasts on-device capabilities with zero cloud dependency.

Coming this month, the system will be available from top manufacturer partners including Acer, ASUS, Dell, Gigabyte, HP, and MSI. The new SKU maintains the GB10 Grace Blackwell Superchip, DGX OS, and full NVIDIA AI software stack found in larger models while offering an accessible starting price point of $4,999.

Hardware Architecture and Capabilities

The DGX Spark combines NVIDIA Grace Blackwell compute, unified memory, NVIDIA ConnectX-7 networking, and an NVIDIA CUDA-accelerated software stack into a single compact system. This setup serves as a complete platform for agents, inference, fine-tuning, data science, and edge development.

With 64GB of unified memory, the platform supports models with up to 100 billion parameters entirely on device. Developers can experiment with custom data and run AI assistants locally without relying on external cloud instances for everyday tasks.

Clustering and Scaling With NVIDIA Sync

When workloads outgrow a single device, two DGX Spark 64GB systems can cluster together using the NVIDIA Sync Cluster Assistant without requiring complex infrastructure setups. Connecting units via their built-in ports pools memory to 128GB, expanding model support up to 200 billion parameters while delivering up to 1.7x performance in testing.

The integrated NVIDIA ConnectX-7 networking facilitates direct communication between nodes over a 200 GbE fabric. The accompanying NVIDIA Sync app automatically detects connected units, validates device configurations, and configures the network so developers can focus on their workflows rather than server administration.

Software Support and Ecosystem Integration

The platform ships ready for agent development from day one, supporting tools like the NVIDIA Agent Toolkit, CUDA-X AI libraries, and Nemotron open models. Popular runtimes including Ollama, vLLM, and PyTorch with CUDA are supported natively out of the box, letting users go from power-on to running models within minutes.

Creator application providers are also adopting the platform. Blender is among the first major creator application providers to support DGX Spark, with a prebuilt, downloadable installer coming soon to help users integrate 3D creation pipelines with local AI processing.

Practical Workloads and Availability

The DGX Spark 64GB is built to handle continuous operations, such as running a coding or research agent around the clock to review code, analyze documents, or execute multistep tasks. Users can run language or image generation models locally while freeing up their everyday laptops and desktops for other creative applications.

Developers seeking agentic AI playbooks and deployment guidelines can visit the official NVIDIA build resources. The DGX Spark 64GB is available starting October 23 through hardware partners including ASUS, Gigabyte, and MSI.

Sources

  • NVIDIA BlogNVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI

Continue chronologically

Related entity coverage