What Is the RTX Spark Superchip?
The RTX Spark superchip is NVIDIA’s first all-in-one processor for Windows PCs that combines a 20-core CPU, a next‑generation RTX GPU, and unified memory into a single package designed to run advanced AI workloads locally and efficiently on consumer devices. Unlike a traditional graphics card, RTX Spark merges an NVIDIA Grace CPU and a Blackwell-generation RTX GPU on one platform, linked by the NVLink‑C2C interconnect for ultra‑fast communication. This design focuses on AI computing Windows users can run on‑device, from language models and autonomous agents to creative tools and high‑resolution video editing. NVIDIA describes this as the foundation of the “personal AI computer,” where instead of opening apps for every task, users can ask AI assistants to handle work directly on their NVIDIA AI Windows PC, reducing dependence on cloud services.
Inside the 20-Core Design: Grace CPU Meets Blackwell GPU
At the heart of the RTX Spark superchip is a 20-core NVIDIA Grace CPU based on the ARM instruction set, co‑designed with partners for efficiency and connectivity, and tightly paired with a Blackwell RTX GPU offering 6,144 CUDA cores and 5th‑generation Tensor Cores with FP4 precision. Together they share up to 128GB of unified memory, so AI models and data no longer need to shuffle between separate pools of system RAM and GPU VRAM. NVIDIA states that RTX Spark can reach up to 1 petaflop of AI compute and run language models with up to 120 billion parameters and 1 million token contexts locally. This 20-core processor configuration is tuned more for AI throughput and multi‑agent workloads than for traditional single‑threaded CPU tasks, which marks a shift from classic PC chip design.
Built with Microsoft for AI-Focused Windows PCs
RTX Spark is developed in close collaboration with Microsoft to fit directly into the Windows ecosystem and its new agent‑driven experience. The chip is designed for Windows on ARM, benefitting from improved native app support and compatibility layers like Prism for x86 software. Microsoft adds new security primitives so Windows‑native AI agents can run safely, while NVIDIA’s OpenShell runtime provides privacy, containment, and policy controls over what agents can access. This means sensitive data can stay on your PC or be masked when cloud access is needed. According to NVIDIA’s CEO Jensen Huang, “the PC is being reinvented” as users shift from launching apps to asking AI agents to complete tasks. RTX Spark superchip PCs are expected from brands such as Microsoft Surface, ASUS, Dell, HP, Lenovo, and MSI starting in the fall.
What RTX Spark Means for AI-Focused Windows Users
For AI-focused Windows users, RTX Spark turns a laptop or desktop into an AI workstation that fits in a slim chassis and runs without constant cloud calls. The chip is built for AI computing Windows workloads: running large language models locally, driving Windows-native agents, and handling creative tasks like 12K 4:2:2 video editing, 4K AI video generation, and complex 3D scene rendering with up to 90GB scenes. While NVIDIA claims it can still run modern AA games at 1440p with 100+ FPS, ray tracing, DLSS, and Reflex, gaming is secondary to AI performance. RTX Spark sits at the center of NVIDIA’s strategy to move beyond data centers and into everyday PCs, giving users a personal AI computer that can execute advanced models, agents like Hermes Agent and OpenClaw, and creative AI tools directly on their NVIDIA AI Windows PC.
