NVIDIA RTX Spark Compact Desktops Launch From ASUS, Dell and MSI With Up to 128GB Unified Memory for Local AI

NVIDIA has announced the availability of its new RTX Spark compact desktops from ASUS, Dell, and MSI, bringing local AI development, model fine-tuning, and AI-powered creative workflows to dedicated Windows PCs. Featuring a 20-core NVIDIA Grace CPU, a Blackwell-based RTX GPU, and up to 128GB of unified memory, the systems are designed for developers and creators who want to run larger AI models without depending entirely on cloud computing.

The RTX Spark desktops deliver up to 1 petaflop of AI performance and feature 6,144 CUDA cores. NVIDIA claims the hardware can run large language models with up to 120 billion parameters locally, render 3D scenes exceeding 90GB in real time, and handle 4:2:2 video editing. The systems also include 10Gb Ethernet connectivity for transferring large AI models, datasets, and creative assets.

One of the main advantages of these compact PCs is their ability to keep AI agents and scheduled workloads running independently of a user’s primary computer. Running supported models locally also allows developers to process code and files on their own systems while reducing cloud API expenses.

NVIDIA is expanding the software ecosystem alongside the new hardware. Developers can use native CUDA support and Linux development tools through Windows Subsystem for Linux, while Hermes Agent, NVIDIA NemoClaw, and Unsloth provide tools for building AI agents and fine-tuning supported models. NVIDIA Switchyard, now available in Hermes, can also route tasks between local and cloud-based AI models, allowing less demanding workloads to run locally.

Creative applications are receiving additional support, too. NVIDIA and Adobe are expanding native Windows on Arm support and GPU acceleration across Premiere, Photoshop, Illustrator, InDesign and Substance 3D Painter. Adobe Premiere receives TensorRT-RTX acceleration for Enhance Speech, Object Mask, Scene Edit Detection and Auto Reframe. Adobe also claims the improvements can deliver more than 50% faster turnaround for everyday H.264 video deliverables.

Blender 5.3 is another important addition to the RTX Spark ecosystem, integrating NVIDIA DLSS 4.5 Ray Reconstruction into the Cycles viewport. This technology improves interactive ray-traced previews by reducing noise and refining lighting, reflections, and materials while artists work.

NVIDIA is also highlighting upcoming enterprise AI improvements. Microsoft plans to bring Windows Server 2028 support to the NVIDIA DGX Station for Windows, powered by the GB300 Grace Blackwell Ultra Desktop Superchip. Meanwhile, Windows ML now supports GGUF models from Hugging Face through the llama.cpp backend, giving developers more options for running local machine learning models.

The new RTX Spark compact desktops are available today from ASUS, Dell and MSI. Acer, GIGABYTE and HP are also preparing additional mini PCs, while Microsoft is working on a Surface RTX Spark Dev Box. NVIDIA has not disclosed detailed retail pricing for the new desktop configurations.

SpecificationDetails
CPU20-core NVIDIA Grace CPU
GPU ArchitectureNVIDIA Blackwell
GPUNVIDIA Blackwell RTX GPU
CUDA Cores6,144
AI PerformanceUp to 1 petaflop
Unified MemoryUp to 128GB
Local AI ModelsUp to 120-billion-parameter LLMs
3D Rendering90GB+ scenes in real time
Video EditingReal-time 4:2:2 video editing
Networking10Gb Ethernet
AI DevelopmentNative CUDA and Windows Subsystem for Linux
AI Frameworks & ToolsHermes Agent, NVIDIA NemoClaw, Unsloth, ComfyUI
Creative SoftwareAdobe Premiere, Photoshop, Illustrator, InDesign, Substance 3D Painter
3D SoftwareBlender 5.3 with DLSS 4.5 Ray Reconstruction
Operating SystemWindows
ManufacturersASUS, Dell, and MSI
Upcoming ManufacturersAcer, GIGABYTE and HP
AvailabilityAvailable now
Jani Dushman
Jani Dushman

I'm Jani, a dedicated Tech Writer and Reviewer at Xiaomitoday. With a passion for exploring and dissecting the latest in technology, my mission is to bring you insightful and comprehensive reviews that empower your decision-making in the fast-evolving world of gadgets and tech.

We will be happy to hear your thoughts

      Leave a reply

      XiaomiToday
      Logo