NVIDIA confirms RTX Spark roadmap with Rubin in 2027 and Rosa Feynman in 2029

RTX Spark gets a multi-generation roadmap

NVIDIA confirms RTX Spark roadmap through Rubin and Rosa Feynman

NVIDIA has confirmed that RTX Spark and DGX Station for Windows are not one-off products. During its GTC Taipei keynote, the company showed a roadmap that places both product lines under a new long-term Windows AI PC and workstation strategy.

The first generation is based on Blackwell and launches in 2026. NVIDIA lists RTX Spark laptops and compact desktops next to DGX Station for Windows. RTX Spark uses Grace Blackwell Spark silicon with LPDDR5X memory, while DGX Station uses Blackwell Ultra-class hardware with HBM3e memory and ConnectX-8 800G networking.

Rubin Spark follows in 2027 and 2028

The roadmap then moves to the Rubin generation. NVIDIA lists Vera Rubin Spark with LPDDR6 memory for the 2027 and 2028 window. The same section also includes Rubin with HBM4, Vera CPU and CX9 1600G networking, showing that NVIDIA plans both Spark and DGX Station-class systems beyond Blackwell.

Source: NVIDIA

The most interesting part of the slide is the Feynman generation. NVIDIA lists Feynman with HBM Next, Rosa CPU, CX10 networking and two Rosa Feynman Spark entries for 2029 and 2030. This is the first time NVIDIA has shown the Rosa Feynman Spark name on a public roadmap.

DGX Station is also moving to Windows

NVIDIA is also bringing DGX Station to Windows. DGX Station for Windows is based on the GB300 Grace Blackwell Ultra Desktop Superchip, combining a Blackwell Ultra GPU with a 72-core Grace CPU through NVLink-C2C. NVIDIA lists up to 748GB of coherent memory and up to 20 petaFLOPS of FP4 performance.

DGX Station for Windows are not gaming PCs. They are aimed at enterprise AI workflows, local agents, model development, data science and high-throughput inference. NVIDIA says it can run AI models with up to 1 trillion parameters locally. The system is expected in Q4 from ASUS, Dell Technologies, GIGABYTE, HP, MSI and Supermicro.

Sources: NVIDIA