Nvidia is expanding its DGX Spark lineup with a new 64GB memory configuration. This version is designed to offer a more accessible entry point for local AI inference, as current dense models like Qwen 3.8 27B can operate within 32GB of RAM.
Systems featuring the 64GB GB10 Superchip from manufacturers including Acer, Asus, Dell, Gigabyte, HP, and MSI will launch on October 23. The starting price for these units is set at $4 999, which is lower than the current market price of $7 000 to $9 000 for 128GB models.
The new 64GB systems retain the same 20-core Arm CPU complex and 273 GB/s shared memory bandwidth as the 128GB versions. They also include the ConnectX 7 RDMA NIC, allowing users to cluster multiple units together to increase total memory capacity and performance.
Nvidia is introducing the Sync Cluster Assistant to automate the process of linking multiple Spark systems. Additionally, the Sync application now features a Model Launcher tool, which automates the downloading and deployment of AI models across one or more connected systems.
