Windows PCs Get RTX Spark Chips With Unified Memory For Local AI Tasks
Microsoft and NVIDIA release RTX Spark SoCs for Windows laptops, combining Arm CPUs with Blackwell GPUs to enable local AI agents.
Key points
- RTX Spark combines up to 20 Arm CPU cores with Blackwell GPU technology on a single chip.
- The system supports up to 128GB of unified LPDDR5x memory running at 9,400 Mbps.
- Microsoft introduces a two-click setup process for installing local AI agents via Windows Store.
Microsoft and NVIDIA launched the RTX Spark system-on-a-chip for Windows laptops last week in San Francisco. The new hardware aims to bring powerful local AI capabilities to Windows PCs, addressing legacy compatibility issues while targeting developers, gamers, and creators.
The RTX Spark SoC integrates up to 20 NVIDIA Grace Arm CPU cores with a Blackwell GPU featuring up to 6,144 CUDA cores. According to Engadget, this configuration includes fifth-generation tensor cores and fourth-generation ray tracing cores. A slightly less powerful variant offers 18 CPU cores and 5,120 GPU cores for broader market coverage.
Technical Specifications and Architecture
The chip design emphasizes high-speed unified memory, offering up to 128GB of LPDDR5x memory running at 9,400 Mbps. This architecture mirrors the approach seen in Apple Silicon, aiming to provide efficient performance for high-end PCs. As Threat Times reported on October 15, the launch targets resolving legacy compatibility problems from previous Arm-based Microsoft devices.
RTX Spark also features a Blackwell video engine with support for AV1 UHQ and 4:2:2 encoding and decoding. This hardware setup allows thin-yet-powerful laptops and small desktops to handle content creation and gaming alongside intensive AI tasks locally.
Local AI Agent Integration
Microsoft focused heavily on optimizing Windows for local AI models and agents. Peter Dawoud, principal program manager at Microsoft, stated that a new Get Started experience allows users to find, install, and launch agents in two clicks. Agents are distributed through the Windows Store, simplifying dependency management and updates.
Hermes, Perplexity, and OpenClaw are the first three officially supported agents at launch. During testing, The Verge noted that setting up Hermes was straightforward, but OpenClaw required an uninstall and reinstall to resolve a local gateway error. Meta’s Muse agent is also expected to arrive on Windows in the future.
What should you do?
IT professionals should assess whether RTX Spark devices meet their organization's software compatibility requirements before deployment. Users interested in local AI should verify that their specific AI agents are supported on the Windows Store. Monitor Microsoft and NVIDIA for further driver updates and agent compatibility lists. Once the vendor provides an update for any reported setup issues, apply it promptly.
What to do and how to stay safe: NVIDIA
- Verify that critical business applications run correctly on Arm-based Windows environments before purchasing RTX Spark laptops.
- Check the Windows Store for compatibility of specific local AI agents required for your workflow.
- Monitor official vendor channels for driver updates and bug fixes related to the new SoC.
- Review memory and core requirements to ensure the specific RTX Spark configuration matches performance needs.
Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
What components make up the RTX Spark SoC?
It combines NVIDIA Grace Arm CPU cores with a Blackwell GPU and unified LPDDR5x memory on a single chip.
How many AI agents are supported at launch?
Three agents are officially supported at launch: Hermes, Perplexity, and OpenClaw.
Does RTX Spark support video encoding?
Yes, it includes a Blackwell video engine with support for AV1 UHQ and 4:2:2 encoding and decoding.
