NVIDIA DGX Spark Hardware Refresh Increases Unified Memory to 64GB
NVIDIA is updating the DGX Spark hardware platform to include 64GB of unified memory. This hardware refresh enables developers to run larger open-source AI models locally on their workstations.
Verified State Diff
Impact & Verification Analysis
AI developers and researchers building and testing local AI agents on workstation-class hardware.
Increased unified memory allows for the local execution of larger, more complex AI models that previously required cloud-based infrastructure or multi-GPU setups, reducing latency and data privacy concerns for local development.
Full Fact Overview
The NVIDIA DGX Spark platform has been upgraded to support 64GB of unified memory, provided through manufacturing partners such as Acer. This increase in memory capacity is specifically designed to accommodate the growing parameter sizes of open-source large language models and AI agents that require significant VRAM for local inference and fine-tuning tasks.