© ROOT-NATION.com - Use of content is permitted with a backlink.
NVIDIA has officially announced the release of the NVIDIA RTX Spark superchip, which aims to fundamentally change the Windows PC market in the era of personal AI agents. The new hardware platform transforms the PC from an ordinary working tool to a full-fledged digital partner. RTX Spark combines 30 years of technological innovation from the manufacturer – including CUDA technologies, DLSS intelligent scaling, FP4 computing, TensorRT and OptiX platforms, as well as Reflex and G-SYNC systems. All this will be available in ultra-thin notebooks with long battery life and in compact, energy-efficient desktop systems.
“The PC is being reborn,” said NVIDIA founder and CEO Jensen Huang, “For forty years, users have simply launched programs with clicks and text. With RTX Spark and Windows, you simply set a task and the computer does the work. This superchip combines local agents, advanced models, creative processes, and RTX-enabled gaming on a single laptop. This is a new personal AI computer.”
TABLE OF CONTENTS:
Specifications and architecture
The superchip is designed with a focus on maximizing the performance of artificial intelligence, gaming, and content creation:
- Graphics subsystem: GPU on NVIDIA Blackwell architecture with 6144 CUDA cores and fifth-generation tensor cores with FP4 computing accuracy
- Central Processing Unit: 20-core NVIDIA Grace CPU, developed in close collaboration with MediaTek. MediaTek’s expertise in Arm architecture has resulted in the best possible energy efficiency and connectivity
- Interconnect: Components are connected by NVIDIA’s ultra-fast NVLink-C2C proprietary bus
- Memory and power: The platform offers up to 1 petaflop of computing power for AI and 128 GB of unified memory.
These parameters allow running large language models of up to 120 billion parameters with a context window of 1 million tokens locally, directly on the device. Users will be able to edit 12K video (4:2:2) and work with three-dimensional scenes up to 90 GB.
Security and ecosystem of local AI agents
The widespread adoption of popular open source AI projects such as OpenClaw and Hermes Agent was previously hampered by privacy concerns in the cloud. To ensure completely private and secure execution of tasks on users’ devices, NVIDIA has teamed up with Microsoft.
Together, they are introducing new security primitives at the Windows level (identification, process isolation, security policies) and the NVIDIA OpenShell runtime. This complex allows the user to clearly define the boundaries of permissions for assistants, intelligently distribute requests between local models, and automatically mask personal data if the request is sent to the cloud. The developers of OpenClaw and Nous Research (creators of Hermes Agent) have already announced the full integration of their Windows applications with these security protocols, which will allow you to securely interact with files, code plug-ins, and automate work between different programs.
For his part, Microsoft CEO Satya Nadella noted that the company’s goal is to bring unlimited intelligence to every Windows workstation, and the RTX Spark superchip is a breakthrough on this path. For large corporate customers, the company offers the DGX Station for Windows, a Blackwell-based desktop supercomputer for AI developers.
Support for professional software and games
The RTX Spark platform takes graphics capabilities to a new level. It supports the new DLSS 4.5 Ray Reconstruction technology based on second-generation transformer neural network models (coming to Blender 5.3 and dozens of games), as well as RTX Video with 4x frame generation for ComfyUI. More than 100 software and game developers are already adapting their products to the new architecture.
Creative software
Adobe: The company is completely rewriting the architecture of Premiere and Photoshop for RTX Spark. Tools based on the Firefly neural network (Generative Fill and Generative Extend) will receive a twofold acceleration. The new Premiere video pipeline uses unified memory and TensorRT to process timelines in real time, and Substance 3D Painter and Stager will work natively. In addition, AI assistants are integrated directly into the interface of Adobe programs as digital assistants
Blackmagic Design and OTOY: Company executives confirmed that thanks to the autonomy of the chip and support for the Render Network for Octane, the new devices will create a new class of portable studios for video production and 3D artists
ComfyUI and llama.cpp: The developers emphasize that the combination of large memory and local power will allow creating complex multimodal media files and running optimized local models with unprecedented speed.
Gaming industry
Gamers will be able to play AAA titles in 1440p resolution with over 100 FPS, ray tracing, and Reflex. The platform is supported by XBOX (providing easy access to services on new PCs), NetEase (for optimizing games such as NARAKA: BLADEPOINT), and Remedy Entertainment.
Availability of premium devices
The RTX Spark notebooks will feature premium aluminum chassis with a thickness of 14 mm and weighing about three pounds (approximately 1.36 kg). They will feature bright and accurate tandem OLED displays with G-SYNC support. Ultra-efficient compact desktop PCs will also be launched.
The first devices from leading brands, including ASUS, Dell (including XPS 16 Creator Edition), HP (OmniBook line), Lenovo, Microsoft Surface (Surface Laptop Ultra) and MSI, will be available this fall. Models from Acer and GIGABYTE will be released a little later. More technical details about the integration of OpenShell and Windows security protocols for developers will be revealed at the Microsoft Build conference, which will be held on June 2-3.
Read also:
- Sensation at Computex 2026: NVIDIA gives away its most powerful AI to humanity for free
- NVIDIA DLSS 4.5 comes to Star Wars: Galactic Racer and 007 First Light