SIAgentNews.com
← All news
DevelopingAgentsChips

NVIDIA launches RTX Spark mini PCs for local AI performance

NVIDIA is releasing compact desktops from partners like ASUS and Dell that can run AI tasks locally with up to 1 petaflop of performance.

· First reported by NeoTeo

1 petafloppeak AI performance

The short version

  • RTX Spark platform features a 20-core Grace CPU and a Blackwell RTX GPU with 6,144 CUDA cores.
  • Systems can support up to 128 GB of unified memory for handling large AI workloads.
  • ASUS ProArt GR1X is one available model, measuring 150 x 150 x 51 mm and weighing 1.48 kg.
  • NVIDIA lists Acer, GIGABYTE, and HP mini PCs as forthcoming hardware options.

Why it matters

This move signals a shift toward bringing high-end AI capabilities into home and small office environments. It allows developers and enthusiasts to run private, powerful AI models without needing constant internet connections or expensive cloud subscriptions.

The RTX Spark launch

NVIDIA officially announced the availability of its RTX Spark compact desktops on November 18, 2026. These small-form-factor machines are designed to provide significant computing power in a much smaller footprint than traditional workstations. The company is working with several major hardware partners to bring these systems to the market.

As reported by NeoTeo, the initial rollout includes compact desktops from manufacturers like ASUS, Dell, and MSI. These devices are built to handle the specific demands of artificial intelligence. By offering these as pre-configured systems, NVIDIA aims to make high-end hardware more accessible to a broader range of users.

While some models are ready for purchase now, NVIDIA also identified several other partners for the platform. The company listed Acer, GIGABYTE, and HP as manufacturers of forthcoming mini PCs. Additionally, Microsoft will offer a Surface RTX Spark Dev Box for those who need a dedicated development environment.

Core hardware specifications

The RTX Spark platform is built around a high-performance architecture that combines several key components. At its center is a 20-core Grace CPU paired with a Blackwell RTX GPU. This GPU contains 6,144 CUDA cores, which are the basic processing units used to perform the heavy math required for AI.

One of the most significant features of the platform is its support for up to 128 GB of unified memory. Unified memory allows the CPU and GPU to share the same pool of memory, which is a critical requirement for running large language models. The system also includes 10-gigabit Ethernet for high-speed networking.

NVIDIA claims that the RTX Spark platform can reach a peak AI performance of up to 1 petaflop. A petaflop is a measurement of performance representing one quadrillion floating-point operations per second. This level of power enables the machine to process complex AI tasks quickly and efficiently in a compact box.

The ASUS ProArt GR1X model

To see how these specifications translate into a physical product, NVIDIA highlights the ASUS ProArt GR1X. This specific compact desktop serves as a tangible example of the RTX Spark platform. ASUS designed the device to be extremely small, with dimensions of 150 x 150 x 51 mm. Despite its tiny size, it weighs 1.48 kg and is built to handle intense workloads.

The ASUS model offers different configurations, including versions with either 64 GB or 128 GB of unified memory. It is equipped with two M.2 NVMe SSD slots that support PCIe 5.0 and 4.0. These fast storage slots ensure that data can be moved quickly between the drive and the processor during heavy use.

For connectivity, the GR1X includes four USB 3.2 Gen 2x2 Type-C ports and HDMI 2.1. It is capable of driving up to four 4K displays, or a single 8K display via HDMI. To manage the heat generated by these components, ASUS uses a dual-fan cooling design with a 140-watt TDP specification.

Running local AI tasks

The practical application of this hardware is demonstrated through the MSI EdgeMesa N AI+ model. In hands-on testing by ETA Prime, the system was shown running local AI tasks using tools like Hermes Agent and ComfyUI. These examples show how users can interact with AI models directly on their own hardware rather than through a web browser.

The demonstration highlighted how the system handles unified-memory allocation through the NVIDIA app. This process allows the software to dynamically assign memory to different AI tasks as needed. It provides a glimpse into how the RTX Spark platform manages the heavy lifting of local inference and generation.

Beyond just AI, the hardware was also tested for standard computing and gaming. The review by ETA Prime walked through the cooling efficiency and general performance metrics. These tests confirm that while the focus is on AI, the machines remain capable of high-end general-purpose computing as well.

The shift to local AI

The release of the RTX Spark platform represents a broader move by NVIDIA toward local-AI computing. By providing compact desktops with high memory and specialized GPUs, NVIDIA is making it easier for people to run models privately. This reduces the need for users to send their data to external servers for processing.

For developers and creators, these machines offer a way to build and test applications with high performance. The inclusion of 10-gigabit Ethernet and high-speed storage makes them suitable for professional environments. It allows for a more seamless workflow where the AI is an integrated part of the local workstation.

As more partners like HP and Acer join the lineup, the variety of form factors will likely increase. Users will be able to choose between different styles of mini PCs while keeping the same underlying RTX Spark performance. This expansion suggests that local AI will become a standard feature in high-end home and office setups.

What to watch

  • Which specific AI models will be most commonly run on these 128 GB unified memory systems?
  • How will the forthcoming HP and Acer mini PCs differ in cooling and design from the ASUS ProArt GR1X?

Try it

Check the official NVIDIA website to see the full list of partners and specific model availability for the RTX Spark platform.

Sources and links

Read the original at NeoTeo

First reported by NeoTeo. 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.

More AI news