- Nvidia DGX Spark runs larger AI models locally using massive 128GB unified memory
- Native CUDA support makes Spark ideal for advanced AI workloads on desktops
- Its Arm CPU and Blackwell GPU combination avoids expensive professional graphics cards
The highly anticipated Nvidia DGX Spark has finally arrived in the form of a very small desktop system built around the GB10 superchip.
It has 128GB of shared LPDDR5X memory, a specification that immediately separates the system from typical desktops and even the most compact workstations.
And according to a first examination of Toms Hardwarethe system only delivers good results when its AI-driven capabilities are fully utilized.
Focus on hardware design and connectivity
The Spark’s hardware design is based on a single package combining an Arm-based CPU and a Blackwell GPU.
This integration allows Nvidia to support larger local models without requiring professional-class graphics cards at extreme costs.
Although Apple and AMD systems offer large shared memory configurations, they do not directly support Nvidia’s software ecosystem, which continues to dominate many AI development streams.
The physical design emphasizes density and airflow rather than visual aesthetics or modular expansion.
With a volume of just over a liter and measuring around 150 x 150 x 50mm, the unit fits comfortably among any modern mini PC, but the similarities mostly end there.
Along with a USB-C power input, the unit provides three 20Gbps USB-C ports with DisplayPort Alternate Mode, an HDMI 2.1a port, and a 10Gb Ethernet connection.
Most notably, it includes two QSFP ports driven by an integrated ConnectX-7 network interface capable of reaching up to 200 Gbps, allowing multiple units to be linked together for distributed computing experiences, a capability rarely associated with a mini PC.
The system runs DGX OS, a custom Ubuntu 24.04 LTS distribution closely aligned with Nvidia’s software stack.
It can operate as a locally connected computer with a monitor and keyboard, or as a headless system accessed remotely via a network.
Nvidia’s Sync utility simplifies remote access from Windows and macOS machines, allowing AI tools to run continuously in the background.
These usage patterns resemble how mobile desktops or shared compute nodes are accessed, rather than how everyday desktops are typically used.
The DGX Spark benefits from a unified 128 GB memory pool with native CUDA support, a rare combination in compact systems designed for local AI work.
This configuration allows larger models to run entirely in memory, thus avoiding frequent data movements between system RAM and GPU memory, and therefore some of the practical limitations seen on discrete GPUs with smaller VRAM pools are reduced.
This same capability also introduces obvious tradeoffs. The price of entry remains high compared to compact desktops, especially for users who don’t run demanding AI workloads on a daily basis.
The system does not support Windows, which restricts software compatibility for users outside of Linux environments.
Its GPU is also unsuitable for gaming or general graphics tasks, reinforcing its narrow scope.
DGX Spark assumes that local AI experimentation is a primary and ongoing requirement, but if it’s not your priority, it loses its practical value.
Follow TechRadar on Google News And add us as your favorite source to get our news, reviews and expert opinions in your feeds. Make sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, unboxings in video form and receive regular updates from us on WhatsApp Also.
