Nvidia launches RTX Spark mini desktop for local AI processing
The new mini PC allows users to run large language models and high-end video editing locally without sending data to the cloud.
The short version
- The device features a 20-core Nvidia Grace CPU and a Blackwell RTX GPU with 6,144 CUDA cores.
- It supports running 120 billion-parameter large language models and real-time 3D scene rendering.
- Adobe Premiere Pro sees a performance increase of over 50 percent for H.264 video production.
- The system includes 'Switchyard' to manage costs by splitting tasks between local hardware and the cloud.
- Preorders for Asus products have already begun in South Korea via Coupang and Compuzone.
Why it matters
This hardware makes it possible for individuals and small businesses to run sophisticated AI models privately on their own desks. It removes the need to send sensitive files to external servers while providing professional-grade speeds for video and 3D work.
A new standard for local AI
Nvidia has officially announced the RTX Spark mini desktop, a compact computer designed to handle intensive artificial intelligence tasks. The hardware is being sold through major partners including Asus, Dell, and MSI. This system is built to manage local AI agent operations, image and video generation, and model fine-tuning all on a single machine.
One of the primary advantages of this hardware is its ability to process data locally. Users can run supported models without ever sending their private code or files outside of the physical device. This provides a layer of privacy and security for those who handle sensitive information but still require high-level AI capabilities.
The hardware is powered by a 20-core Nvidia Grace CPU and a Blackwell RTX GPU. This GPU contains 6,144 CUDA cores, which are the basic processing units used to accelerate the complex math required for AI. The system also features up to 128GB of unified memory and can reach AI performance levels of up to 1 petaflop.
Powering heavy creative workflows
Beyond general AI tasks, the RTX Spark is designed to significantly improve professional creative software. For example, Adobe has added support for RTX acceleration and Windows on Arm native support. This allows the Adobe Premiere Pro application to be more than 50 percent faster when producing H.264 video content.
The performance gains extend to other creative tools as well. Users of Adobe Illustrator can see up to 30 percent higher performance compared to using emulation. Additionally, Blender 5.3 now includes DLSS 4.5 ray reconstruction in the Cycles viewport, which helps render realistic lighting and reflections in 3D scenes.
The machine is also capable of handling massive data loads in real time. It can manage the real-time rendering of 3D scenes that are larger than 90GB. It also supports real-time editing of 4:2:2 video, which is a high-quality video format often used by professional filmmakers and content creators.
Tools for developers and users
Developers have several native options for building workflows on this hardware. They can use native CUDA and the Windows Subsystem for Linux to create local AI environments. The system also provides specific blueprints, such as 'Hermes Agent' or 'NemoClo,' to help users set up their own AI agents quickly.
For those who need to fine-tune models, the system supports the 'Unsloth' framework. This allows users to customize existing AI models to perform specific tasks more accurately. These tools are designed to make it easier for people who are not engineers to deploy functional AI systems on their own desks.
The hardware also includes a feature called 'Switchyard' to help manage operational costs. This software can split tasks between the local device and the cloud. It allows routine work to run on the local RTX Spark hardware while reserving more advanced reasoning tasks for cloud models, which can save on expensive cloud computing fees.
Availability and product lineup
The rollout of the RTX Spark mini desktop follows the recent release of the RTX Spark laptops. In South Korea, preorders for Asus products have already started on the retail platforms Coupang and Compuzone. The official release dates for products from Asus and MSI are scheduled between late November and early December.
Other partners will begin releasing their versions of the hardware sequentially following the initial launch. This indicates a broad hardware push from Nvidia to put high-end AI capabilities into the hands of more diverse users. The release schedule ensures that various brands will have the hardware available for different market segments.
The lineup is expected to expand further in the coming months. Future releases will include mini PCs from Acer, Gigabyte, and HP. Additionally, Microsoft is set to release a 'Surface RTX Spark Dev Box,' which will likely cater to developers who need a standardized environment for testing and building AI applications.
What to watch
- How the 'Switchyard' cost-splitting feature performs in real-world cloud-hybrid scenarios.
- The release dates for the Microsoft Surface RTX Spark Dev Box.
Try it
Check the Coupang or Compuzone websites if you are in South Korea to see the current pre-order status for the Asus RTX Spark.
Sources and links
- Reporting디지털투데이: the article we sourced
First reported by 디지털투데이. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.