NVIDIA DGX Spark - A Grace Blackwell AI Supercomputer on Your Desk

NVIDIA DGX Spark

A Grace Blackwell AI Supercomputer on Your Desk

BUY NOW

A Grace Blackwell AI supercomputer on your desk.

NVIDIA DGX™ Spark belongs to a new class of computers designed to build and run AI, delivering up to 1,000 AI TOPS performance from a power-efficient, compact form factor. With the preinstalled NVIDIA AI software stack and 128GB of memory, developers can prototype, fine-tune, and inference large AI models of up to 200 billion parameters locally and seamlessly deploy to the data center or cloud.

NVIDIA Spark Diagram
FEATURES

NVIDIA GPU, CPU, Networking, and AI Software Technologies

NVIDIA GB10 Superchip
NVIDIA GB10 Superchip

Experience up to 1000 AI TOPS of AI performance at FP4 precision with the NVIDIA Grace Blackwell architecture.

128GB Coherent System Memory
128 GB of Coherent Unified System Memory

Run AI development and testing workloads with AI models up to 200 billion parameters at your desktop with large, unified system memory.

NVIDIA ConnectX Networking
NVIDIA ConnectX Networking

High-performance NVIDIA Connect-X networking enables connecting two NVIDIA DGX Spark systems together to work with AI models up to 405 billion parameters.

NVIDIA AI Software Stack
NVIDIA AI Software Stack

Use a full-stack solution for generative AI workloads, including tools, frameworks, libraries, and pretrained models.

DGX OS

NVIDIA DGX OS provides a customized installation of Ubuntu Linux with system-specific optimizations and configurations, and additional drivers. It provides a stable, fully-tested, and supported OS to run AI, machine learning, and analytics applications on DGX Supercomputers.

Key features include:
  • DGX system-specific performance optimizations
  • NVIDIA GPU driver, CUDA toolkit, domain specific libraries
  • Docker Engine
  • NVIDIA Container Toolkit
  • ConnectX networking drivers
NVIDIA DGX OS running on laptop

200B Parameters

Work with AI models and workloads that can’t run on existing laptop or desktop systems with 128GB of Unified memory allowing up to 200B parameter models.

NVIDIA DGX 200B Parameters

ConnectX-7

Connect two DGX Spark systems together via ConnectX-7 to work with models of up to 405 billion parameters.

NVIDIA DGX ConnectX-7

Data Security

NVIDIA DGX Spark is a single user system offering powerful AI compute for large models with no need to transfer data to external locations.

NVIDIA DGX Security

DGX Spark in Scan Cloud

NVIDIA DGX Spark Cloud Laptop

Accelerate Innovation with DGX Spark in Scan Cloud

DGX Spark in Scan Cloud empowers teams to rapidly build, test, and finetune AI workloads in a private, secure, and scalable lab environment. Whether you're finetuning models, or experimenting with larger models thanks to DGX Spark's 128GB of memory, DGX Spark in Scan Cloud offers a scalable platform that can burst models from Spark into enterprise cards and even DGX appliances.

From Lab to Enterprise - Start small, scale smart.

DGX Spark is uniquely integrated with Scan’s enterprise-grade GPU cards and full-scale DGX appliances, enabling you to move from development to deployment without rearchitecting your workloads. What begins in a development lab environment can scale effortlessly into production with the unmatched performance of NVIDIA DGX systems—perfect for demanding AI, ML, and data science pipelines.

NVIDIA DGX Spark Cloud Server Racks
NVIDIA DGX SPARK Workloads

DGX Spark Comparison

DGX Spark* PC Workstation Data Center Cloud
AI Performance (Single GPU) 1,000 TOPS Up to 3,352 TOPS (GeForce 5090) Up to 4,000 TOPS (RTX PRO 5000) Variable Variable
Maximum Memory (Single GPU) 128GB Up to 32GB (GeForce 5090) Up to 96GB (RTX PRO 5000) 141GB (H100) 141GB (H100)
Maximum Memory (Multi-GPU) 2x Spark: 256GB 2x GPUs: 64GB 4x GPUs: 384GB Variable Variable
Max AI Model (4-bit) (Single GPU) 200B** 51B** 153B** Up to Very Large Up to Very Large
Max AI Model (4-bit) (Multi-GPU) 2x 400B** (2000 combined TOPS) 2x 102B** (6,700 combined TOPS w/o 5090 GPUs***) 4x 614B (16,000 combined TOPS w/4x RTX PRO 5000 GPUs) Up to Very Large (very large TOPS) Up to Very Large (very large TOPS)
*Preliminary DGX Spark Data, specifications subject to change
** Assuming 4-bit precision, 20% overhead for model
*** Assuming 2x 5090 GPUs can fit into a single system
SHOP NOW SHOP NOW SHOP NOW LET'S CHAT LEARN MORE
Read the Technical Comparison
Read Article

DGX Spark Performance Comparison

The DGX Spark is a powerful, affordable desktop device built for developing advanced AI models like generative and reasoning AI. It's a great fit for many use cases—but the best setup always depends on your workload. If your training models require less memory and more performance then its possible there may be a better solution for you. Speak to our experts to find the right solution for you.

Let's Chat
TOPS Performance
Model parameters
RTX 5090 PC
3300
51B
RTX PRO Workstation
4000
153B
1x DGX Spark
1000
200B
2x RTX PRO Workstation
8000
307B
2x DGX Spark
2000
400B
4x RTX PRO Workstation
13000
614B
Data Centre / Cloud
1,000,000+
Really Large

Frequently Asked Questions (FAQ)

NVIDIA DGX Spark belongs to a new class of computers created for the AI era. Designed from the ground up to build and run AI, DGX Spark is an ideal platform for AI developers who often need to work on public or private clouds due to insufficient memory or lack of access to the optimal software stack their workloads require.

  • Powered by the GB10 Superchip
  • 1,000 TOPS of AI performance using FP4
  • NVIDIA Blackwell GPU with fifth-generation Tensor Core technology
  • NVIDIA Grace CPU implementation with 20-core high-performance Arm architecture
  • 128 GB of coherent, unified memory
  • Up to 4 TB of NVMe storage
  • Support for up to 200B-parameter large language models
  • Connectivity with NVIDIA Connect-X networking to link two Project DIGITS units for up to 405B-parameter models
  • Compact design, takes up minimal desktop space
  • Plugs into standard wall outlet

Preliminary, subject to change:

  • Blackwell GPU - Up to 1,000 AI TOPS FP4
  • Tensor Cores: 5th Generation
  • RT Cores: 4th Generation
  • CUDA Cores: Blackwell Generation
  • Grace CPU: 10 performance cores (P-Cores) Cortex-X925 + 10 efficiency cores (E-Cores) Cortex-A725
  • NVLink-C2C (chip to chip interconnect) 5x the bandwidth of PCIe 5
  • Memory: 128GB of coherent, unified LPDDR5X Low-Power Memory
  • Memory bandwidth: 273 GB/s
  • Storage: Up to 4TB SSD storage
  • Network connectivity: 10-GbE (RJ-45) and Connect-X (x2 QSFP)
  • Display: HDMI 2.1
  • NIC: ConnectX-7 SmartNIC
  • USB: 4x USB Type C
  • WiFi: WiFi-7
  • Bluetooth: BT 5.3
  • Audio output: HDMI multichannel audio output
  • NVENC | NVDEC: 1x | 1x
  • OS: NVIDIA DGX OS
  • System dimensions: 150mm L x 150mm W x 50.5mm H
  • Weight: 1.2kg

DGX Spark can be powered from a standard wall outlet.

The GB10 Superchip is a system-on-a-chip based on a Grace Blackwell architecture with 128GB of unified memory. It provides up to one petaflop of AI performance at FP4 precision and features a NVIDIA Blackwell GPU with the latest fifth-generation Tensor Core and RT Core technology, along with a high-performance Grace CPU built with 20 Arm cores.

The GB10 architecture allows enterprises and researchers to prototype, fine-tune, and test models on local DGX Spark systems and deploy in a data center or cloud. This enables efficient fine-tuning and inference workloads while maintaining consistency across development and deployment environments.

GB10 has ten Cortex-X925 and ten Cortex-A725 cores.

The DGX Spark is designed for AI development workloads such as prototyping, model fine-tuning, inference, and data science. DGX Spark systems can be used in a standalone configuration with display, keyboard, and mouse directly connected, or it can be used in headless configuration as a network connected compute device with no directly connected display, mouse or keyboard.

A 1 PetaFLOP FP4 Blackwell GPU.

DGX Spark ships with DGX OS and additional NVIDIA tools. DGX OS is based on Ubuntu and is consistent with the Linux distribution running on other DGX platforms.

DGX Spark comes pre-installed and/or supports a large catalog of tools, frameworks, models and software for development use.

DGX Spark users can register free of charge for the NVIDIA Developer Program and NVIDIA NGC to get access to SDKs, toolkits, trained models, workflows, and much more.

DGX Spark is equipped with 128 GB of coherent, unified LPDDR5X memory and up to 4 TB of NVMe storage.

Two DGX Spark systems can be connected to offer 256GB or memory and up to 8TB total storage.

DGX Spark runs DGX OS, based on Ubuntu and is consistent with the OS distributions running on DGX platforms.

Users can join the NVIDIA Developer Program and get access to extensive software support, documentation and the developer community for rich community-based support.

Hardware support will be provided by the vendor.

Customers are welcome to choose between the options available for sale. We do not support upgrades to DGX Spark at this time. Any attempt to tamper with or change the components will void the warranty.

We support connecting of up to 2 DGX Spark with an approved QSFP cable at this time.

scan computers ai department logo

01204 474210

Contact our AI team

Call us on 01204 474210