Nvidia introduces RTX Spark platform

Nvidia has announced the RTX Spark, a new PC platform designed for local AI and creators. The platform, set to launch on October 16, 2026, features an ARM-based processor and a Blackwell GPU architecture, targeting developers and content creators who use AI tools extensively.

RTX Spark's hardware specifications

The platform is powered by the N1X chip, which combines a 20-core ARM processor and a GPU based on the Blackwell architecture. It offers unified memory capacity ranging from 24 to 128 GB, manufactured on a 3 nm process by TSMC. The chip includes 10 high-performance Cortex X925 cores and 10 efficient Cortex A725 cores, along with 6144 CUDA units and advanced AI features like fourth-generation RT cores and fifth-generation Tensor Cores.

Compatibility with Windows on ARM

The N1X chip is designed to work with Microsoft's Windows on ARM ecosystem, which has seen improvements in recent years. Native support for major applications like those from Adobe has been added. However, the platform may face performance challenges in gaming due to the unified memory bandwidth limit of 300 GB/s, which is lower than the 672 GB/s of a conventional RTX 5070.

Nvidia's partnerships and configurations

Nvidia partnered with MediaTek for the CPU portion of the N1X chip, as it does not have a license for x86 architecture. The chip is available in three configurations, including a high-end version with up to 128 GB of memory, a mid-range version with 5120 CUDA units, and a mini-PC version with higher thermal design power. The platform will be available through partners like Microsoft, Asus, Lenovo, MSI, and HP, starting at 2899 euros for the Surface Laptop Ultra with 24 GB of memory.

Support for gaming and AI workloads

The RTX Spark platform supports both gaming and AI workloads with features like DLSS 5, Frame Generation, and Nvidia Broadcast. However, gaming performance may lag due to unified memory limitations and the need for Windows emulation for x86 games. The platform also includes local AI capabilities, allowing users to run large AI models without relying on cloud services, supported by the Blackwell GPU's NVFP4 format.

Microsoft collaboration and software integration

Microsoft has been working with Nvidia to adapt Windows to the new architecture, including changes to memory management and workload scheduling. Windows is expected to include features to manage AI agents, preventing them from interfering with the user experience. Nvidia is also integrating its software with Microsoft's Execution Containers (MXC) and plans to support open-source agents like Hermes.

Future outlook for RTX Spark

The RTX Spark platform represents a significant step in the evolution of local AI and computing, aiming to balance gaming, creation, and mobility. Its success will depend on performance, compatibility, and software support, which will be evaluated through detailed testing and benchmarking in the coming days.