Nvidia presents RTX Spark, an Arm platform for Windows laptops focused on local AI
🎧 Listen to this story
Nvidia presented the RTX Spark platform for Arm-based Windows laptops, combining a Grace CPU, RTX Blackwell GPU, and unified memory. According to Xataka, the N1X chip can have up to 20 CPU cores, 6,144 GPU cores, 1 PFLOP of FP4 performance, and up to 128GB of memory with 300GB/s of bandwidth; demonstrations included games such as Cyberpunk and CONTROL Resonant, as well as local AI model execution. The publication notes that these figures and results were part of Nvidia’s presentation.
The Verge reports that Microsoft created a Windows setup experience capable of installing agents in two clicks, with Hermes, Perplexity, and OpenClaw officially supported at launch. Agents can be installed and updated through the Windows Store, while the Prism emulator runs apps without a native Arm version. In the outlet’s testing, setting up OpenClaw required a reinstall, and large local models responded more slowly than cloud models.
Why it matters · editorial interpretation
For those evaluating Windows laptops for local AI, the proposal combines hardware capacity with a simpler way to install agents. However, The Verge’s testing indicates that the experience may still involve reinstalls and slower responses from local models than from cloud models.
Sources
Hardware