NVIDIA DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI
nvidia ·
NVIDIA is expanding its local AI capabilities with the new DGX Spark 64GB configuration, available from major partners like Acer, ASUS, Dell, and HP starting October 23. This compact supercomputer features the GB10 Grace Blackwell Superchip and unified memory, allowing developers to run models up to 100 billion parameters entirely on-device without cloud dependency. The system comes pre-configured with DGX OS and a full NVIDIA AI software stack, supporting popular frameworks such as Ollama, vLLM, and PyTorch out of the box. For scaling workloads, two units can be clustered together using the NVIDIA Sync Cluster Assistant via ConnectX-7 networking, effectively pooling memory to 128GB and boosting performance by up to 1.7x. This clustering capability enables the handling of larger models or concurrent agent requests while maintaining seamless software compatibility across nodes. Additionally, upcoming tools like the NVIDIA Sync Model Launcher aim to simplify the deployment of local AI applications for developers. The device starts at $4,999, positioning itself as an accessible entry point for personal AI supercomputing and edge development.
nvidia