Small Box. Big AI.
NVIDIA DGX Spark
Powered by the NVIDIA GB10 Grace Blackwell Superchip, NVIDIA DGX™ Spark delivers 1 petaFLOP of AI performance in a compact, power-efficient form factor. With 128GB of coherent unified memory and the NVIDIA AI software stack preinstalled, developers can prototype, fine-tune, and run large-scale reasoning models with up to 200 billion parameters locally, then deploy to the data center or cloud.
NVIDIA GPU, CPU, Networking, and AI Software
NVIDIA GB10 Superchip
Get 1 petaFLOP of AI perforamnce at FP4 precision, powered by the NVIDIA Grace Blackwell architecture.
128 GB Coherent Unified System Memory
Develop and test AI models with up to 200 billion parameters directly on your desktop using large coherent unified system memory.
NVIDIA ConnectX Networking
High-speed NVIDIA ConnectX®-7 networking enables two NVIDIA DGX Spark systems to work together, supporting AI models up to 405 billion parameters.
NVIDIA AI Software Stack
Access a complete software stack for generative AI, including frameworks, libraries, tools, and pretrained models.
Engineered for AI at Scale
NVIDIA DGX Spark is powered by the GB10 Superchip, combining a Blackwell GPU with fifth-generation Tensor Cores and FP4 precision for 1 petaFLOP of AI performance. Its 20-core Grace CPU accelerates data prep and inference, while NVLink®-C2C enables a coherent unified memory model with 5X the bandwidth of PCIe Gen 5.
Cluster Two for 405B Models
NVIDIA DGX Spark systems include NVIDIA ConnectX® networking, enabling developers to connect two systems for inference and fine-tuning of AI models up to 405 billion parameters. This configuration is optimized for local-scale deployments of large models like Llama 3.1, helping developers accelerate iteration without the cloud.
With shared architecture across the DGX platform, workloads built on NVIDIA DGX Spark can be deployed seamlessly to the data center or NVIDIA DGX™ Cloud when needed.
AMAX Powers Your AI Journey
AMAX is an NVIDIA DGX AI Compute Systems Elite Partner, the highest level of DGX competency and infrastructure delivery.
We deliver AI infrastructure that grows with you. Start on-prem with NVIDIA DGX Spark and scale seamlessly to NVIDIA DGX BasePOD™, NVIDIA DGX Super POD™ or any NVIDIA accelerated architecture.
From desktop to data center, AMAX provides the full-stack expertise to build, deploy, and expand your AI strategy with confidence.
NVIDIA DGX™ Spark
Built on the NVIDIA GB10 Grace Blackwell Superchip, NVIDIA DGX Spark delivers 1 petaFLOP of AI performance and 128GB coherent unified memory in a compact system designed for local development of large AI models.
- Grace Blackwell Superchip
- 128GB coherent unified system memory
- Supports models up to 200B parameters
- Cluster two systems to run up to 405B parameter models