NVIDIA DGX Spark AI Supercomputer Bundle (3x Systems)

Triple AI Power for Enterprise-Scale Workloads

The NVIDIA DGX Spark Bundle (3x Systems) is a powerful AI computing solution designed for organizations that require high-performance, scalable infrastructure. This bundle includes three DGX Spark units connected via high-speed QSFP interconnect cables, enabling distributed AI processing and significantly enhanced performance.

Built on the advanced NVIDIA Grace Blackwell architecture, each DGX Spark system delivers up to 1 PetaFLOP of AI performance, meaning this triple-node configuration can achieve massive parallel compute power for training, inference, and large-scale AI model deployment.

Key Features & Highlights

Triple-System AI Performance

✅ Includes 3× DGX Spark AI supercomputers + 3× QSFP interconnect cables
✅ Up to 3 PetaFLOPS combined AI performance (FP4)
✅ Designed for distributed AI training and clustering

Massive Unified Memory for Large Models

✅ 384GB LPDDR5x unified memory (128GB per system)
✅ Coherent CPU-GPU memory architecture for faster processing
✅ Supports extremely large AI models and datasets locally

Each DGX Spark system uses unified memory shared between CPU and GPU, eliminating bottlenecks and improving performance for AI workloads.

Grace Blackwell Superchip Architecture

✅ Each unit powered by:

✅20-core Arm CPU

✅NVIDIA Blackwell GPU with 5th Gen Tensor Cores

✅ High-speed NVLink-C2C interconnect for seamless compute

This architecture enables efficient data transfer and faster AI computation compared to traditional systems.

Enterprise Storage & High-Speed Access

✅ 12TB total NVMe SSD storage (4TB per system)
✅ Self-encrypting drives for data security
✅ Ideal for large datasets, AI models, and pipelines

High-Speed Networking & Clustering

✅ Includes QSFP high-speed interconnect cables
✅ ConnectX-7 Smart NIC with up to 200Gbps bandwidth
✅ Enables clustering of multiple systems for scalable AI

The integrated networking allows DGX Spark systems to be connected into a high-performance cluster for distributed workloads.

Workflows, Models & AI Tools

The NVIDIA DGX Spark is built to accelerate AI development from day one, providing a complete ecosystem of software, tools, and frameworks required to run advanced workloads seamlessly. Whether you’re building machine learning models, deploying AI applications, or experimenting with generative AI, DGX Spark offers a ready-to-use environment designed for speed and efficiency.

Developers have the flexibility to install and run their preferred frameworks and toolchains, while also benefiting from NVIDIA’s optimized AI ecosystem. This includes access to advanced solutions like NVIDIA NIM, which simplifies model deployment, as well as AI Blueprints for rapid solution development and AI Workbench for streamlined experimentation and workflow management.

Built for Enterprise AI Development

The DGX Spark 3x Bundle is more than just hardware—it is a Complete AI platform with pre-installed NVIDIA AI software stack, including:

✅ CUDA, cuDNN, TensorRT
✅ NVIDIA AI Workbench
✅ Containerized AI workflows

This enables teams to:

✅ Train and fine-tune large language models
✅ Run advanced inference workloads
✅ Deploy AI applications at scale
✅ Build production-ready AI solutions