NVIDIADGX Spark

NVIDIA DGX Spark and other GB10-powered systems are compact AI computers designed for local AI development, fine-tuning, inference, and data science workloads.

The 128GB systems featured on this page use the NVIDIA GB10 Grace Blackwell Superchip and unified memory. Differences between brands mainly include storage, cooling, ports, design, warranty, and support. The product listings distinguish available systems from backorder options.

These are specialised AI systems, not conventional gaming PCs or standard Windows desktops.

NVIDIA's product page lists inference support for models up to approximately 200B parameters on a single 128GB DGX Spark, with fine-tuning support for models up to approximately 70B parameters.

Two compatible 128GB systems provide 256GB across the devices and can support inference for models up to approximately 400B parameters. Four 128GB systems provide 512GB across the devices, with inference support up to approximately 700B parameters in supported configurations.

Actual capacity depends on model precision, quantization, context length and the software used. Each system retains its own memory; compatible networking and distributed software are required.

Two compatible systems can connect using suitable QSFP networking cables, which we sell separately. Four-system clusters require a compatible network switch. Contact us for hardware, cable and multi-system enquiries.

Yes. DGX Spark is designed to run supported AI inference, fine-tuning, prototyping, data science, and development workloads locally. This allows organisations to process suitable models and datasets on their own system instead of relying entirely on cloud-based computing.

Internet access may still be required to download models, containers, software packages, updates, licences, or external services. Certain large-scale training and production workloads may also require additional systems, cloud infrastructure, or data-centre resources.

DGX Spark is built around an Arm-based NVIDIA Grace CPU and is primarily designed to operate with the NVIDIA DGX software environment. Applications, libraries, containers, and development tools must be compatible with the system's operating system and Arm64 architecture.

It should therefore not be considered a direct replacement for a standard Windows x86 desktop or gaming PC. Customers should confirm that their required AI frameworks, applications, containers, and dependencies support the selected DGX Spark or GB10-powered system before purchasing.

IdealTech supplies the DGX Spark or NVIDIA GB10-powered hardware system according to the specifications, accessories, software, and warranty coverage stated in the individual product listing.

AI model development, dataset preparation, model training, application development, API integration, workflow configuration, software migration, and custom AI deployment services are not included unless they are specifically stated or separately quoted.

Customers are advised to confirm their intended models, software frameworks, technical requirements, and deployment environment before placing an order. Our team can assist with product specification and hardware-related enquiries.