Nvidia Prepares RTX Spark N1X Superchips for Fall Windows PC Launch

First unveiled at NVIDIA GTC Taipei on June 1, 2026 (Taipei time), the Nvidia RTX Spark platform is gearing up for its official hardware rollout this fall. Developed in deep collaboration with Microsoft and featuring custom Arm CPU silicon co-designed with MediaTek, the RTX Spark N1X superchip unites power-efficient CPU processing, Blackwell graphics, and fast unified memory to create dedicated hardware for local agentic AI on Windows.

The platform aims to transform Windows PCs into responsive local AI workstations capable of running persistent personal AI agents alongside demanding creative workloads and PC gaming.

A Unified Architecture for Agentic Computing

At the core of the RTX Spark platform is the N1X superchip, manufactured on TSMC's 3-nanometer process node. Rather than relying on traditional PCIe interconnects between a discrete CPU and GPU, RTX Spark unites both processors across a low-latency Nvidia NVLink Chip-to-Chip (C2C) interconnect.

Key Architectural Specifications:

Processor Configurations: The platform offers two main silicon tiers. The flagship configuration features a 20-core Grace Arm CPU (10 performance Cortex-X925 cores and 10 efficiency Cortex-A725 cores) alongside a Blackwell GPU with 6,144 CUDA cores and 24GB to 128GB of unified LPDDR5X memory. A mobile-focused variation pairs an 18-core Arm CPU with a 5,120-core Blackwell GPU and 24GB to 32GB of unified memory.
AI Compute Throughput: Up to 1 petaflop of tensor AI compute power utilizing 5th-generation Tensor Cores with FP4 precision.
Unified Memory Subsystem: Up to 128GB of LPDDR5X memory shared between the CPU and GPU with up to 300 GB/s of bandwidth.
Chassis Design: Engineered for thin-and-light laptop form factors as slim as 14 millimeters and as light as three pounds in 14- to 16-inch sizes.

By sharing a single unified memory pool, large multimodal models and reasoning agents can stay resident in memory without hitting disk swap bottlenecks.

Windows 11 Platform Integration

To take full advantage of the superchip architecture, Microsoft worked alongside Nvidia to implement operating system and scheduler enhancements within Windows 11.

Layer	Components
User space	Autonomous agents (Hermes Agent, OpenClaw); creative tools (Adobe, DaVinci Resolve, Blender); gaming (native Arm64 and Prism-emulated titles)
Runtime	Windows security primitives and Nvidia OpenShell sandboxing; CUDA 13.4 and the TensorRT execution engine; Prism x86/x64 emulation tuned for the RTX Spark microarchitecture
Operating system and hardware control	Workload Profile Scheduling across the 20 heterogeneous cores; Microsoft Power and Thermal Framework (MPTF); unified memory management with raised GPU allocation limits

1. Workload Profile Scheduling (WPS)

To maximize performance on RTX Spark's heterogeneous core layout, Microsoft implemented Workload Profile Scheduling (WPS) within Windows 11. This scheduling enhancement enables the operating system to dynamically distribute compute tasks and scale workloads efficiently across all 20 performance and efficiency cores.

2. Unified Memory Management

Microsoft updated the Windows memory subsystem with a higher, smarter limit on total system memory accessible by the graphics processor. On high-memory systems equipped with up to 128GB of unified memory, this higher allocation ceiling unlocks the ability for local applications to load much larger AI foundation models and render complex creative scenes without memory bottlenecks.

3. Nvidia OpenShell and Agent Privacy Governance

RTX Spark introduces support for the Nvidia OpenShell runtime, built on new Windows security and containment primitives to provide guardrails for local autonomous agents. OpenShell provides user-configurable policy controls to define what agents can and cannot do, intelligently routes queries to local models based on privacy rules, and disguises personal information before queries reach cloud services. Agent developers including Hermes Agent and OpenClaw are integrating both Windows containment primitives and OpenShell for their new applications.

Software Compatibility and Gaming

A key focus for the RTX Spark platform is ensuring broad software and gaming compatibility across Windows on Arm:

Native Creative Applications: Native Arm64 performance is ready for core creative suites, including Adobe Photoshop, Adobe Premiere Pro, Blackmagic DaVinci Resolve, and Blender. Technical tools like MATLAB now officially support Windows on Arm via Prism emulation.
AI Tooling & Developer Roadmap: Microsoft and Nvidia plan to bring CUDA-accelerated PyTorch, llama.cpp, TensorRT, Hugging Face frameworks, Unsloth, and Kohya directly to RTX Spark. Windows ML is also being expanded to enable native TensorRT acceleration in Windows.
CUDA on Windows Arm: CUDA 13.4 introduces official support for Windows on Arm on RTX Spark devices, with display driver 616.41 or later required.
Prism Emulation: Microsoft's Prism translation engine, which previously added support for AVX and AVX2 instruction set extensions, has been tuned for the microarchitecture of RTX Spark to ensure broad compatibility with legacy x86 and x64 applications.
PC Gaming Catalog: Players will gain access to an expanding Windows PC gaming catalog backed by Epic's Easy Anti-Cheat and BattlEye support alongside the Xbox PC app. Riot Games has confirmed that League of Legends and VALORANT are coming to the platform, while PUBG: Battlegrounds will be joining an expansive catalog of compatible titles that already includes Alan Wake 2, Naraka: Bladepoint, and War Thunder.

OEM Hardware Rollout in Fall 2026

According to reporting from Tom's Hardware at Computex, over 30 laptop models and 10 or so desktop configurations are expected to lead the initial hardware rollout starting this fall.

Manufacturer	Flagship Models	Form Factor & Key Features
Microsoft	Surface Laptop Ultra, Surface RTX Spark Dev Box	Laptop Ultra: 15-inch mini-LED PixelSense Ultra touchscreen, maker-friendly ports (USB-C, USB-A, HDMI, SD card reader); Dev Box: compact headless mini-PC with USB-C, USB-A, HDMI, and Ethernet
Dell	XPS 16 Creator Edition	Tandem OLED display (True Black HDR 600), up to 128GB unified memory
ASUS	ProArt P16, ProArt P14	ProArt P16 with up to 4K 120Hz VRR Lumina Pro OLED, ProArt P14 with up to 3K Lumina Pro OLED, ultra-thin thermal modules
HP	OmniBook Ultra 16, OmniBook X 14	OmniBook Ultra 16: 16-inch 3K OLED (DisplayHDR True Black 1000), 99Wh battery with fast charging (up to 50% in ~30 min), quad speakers with smart amps; OmniBook X 14: fast charging with 140W USB-C GaN adapter
Lenovo	Yoga Pro 9n, Yoga 9n 2-in-1	Yoga Pro 9n: clamshell with 15.3-inch 2560x1600 PureSight Pro OLED and up to 80W TDP; Yoga 9n 2-in-1: 16-inch 2880x1800 PureSight Pro OLED convertible
MSI	Prestige N16 Flip AI+	UHD+ Tandem OLED panel, 99.9Wh high-capacity battery

Nvidia notes that additional designs from Acer and Gigabyte will follow the initial launch wave. Among these, Acer has previewed its SFF RTX Spark compact desktop alongside Lenovo systems at IFA.

Outlook

The launch of the RTX Spark superchip represents a significant milestone for Windows computing. By consolidating high-throughput graphics, vector-optimized Arm processing, and massive unified memory onto a high-performance superchip, Nvidia and Microsoft are laying the groundwork for a personal computing landscape where persistent, privacy-preserving AI agents operate smoothly on local hardware.

According to reporting from Tom's Hardware, the first N1X-powered laptops are scheduled to arrive in October 2026 as part of Nvidia's broader fall rollout.