MSI EdgeXpert
Personal AI Supercomputer Based on NVIDIA DGX™ Spark GB10 Platform for Deep Learning
Next-Level AI Power,
Right at Your Desk
The MSI EdgeXpert AI Supercomputer redefines desktop AI computing, delivering petaflop-scale performance through the cutting-edge NVIDIA® GB10 Grace Blackwell Superchip—the same powerhouse at the core of NVIDIA DGX™ Spark. Purpose-built for developers, AI researchers, and data scientists, the EdgeXpert empowers local AI development with unmatched performance, scalability, and advanced features—all in a compact, desktop-ready form.
NVIDIA® Grace Blackwell Architecture
- NVIDIA Blackwell GPU & Arm 20 core CPU Optimizes data preprocessing and orchestration to accelerate model tuning and enable real-time inference with greater efficiency.
- NVLink®-C2C Technology Offers a seamless CPU+GPU memory model with up to five times the bandwidth of PCIe 5.0, ensuring ultra-fast data access and transfer.
Unmatched AI Computing Power
- 1000 AI TOPS (FP4) Tensor Performance Delivers blazing-fast performance for effortlessly running complex AI workloads at scale.
- 128 GB LPDDR5x, unified system memory Provides the large memory needed for smooth model development, rapid experimentation, and high-efficiency inference.
Running and scaling LLMs locally, from 200B to 405B parameters
- Supports AI models with up to 200 billion parameters Run LLMs locally for data security, low latency, cost control, and full AI workflow—prototyping, fine-tuning, and inference.
- Stack via NVIDIA ConnectX: Large AI models & Performance High-performance NVIDIA ConnectX networking enables two MSI systems together to work with AI models up to 405 billion parameters.
Seamless AI Model Scaling from Desktop to Cloud
Leverage NVIDIA's AI software architecture to seamlessly scale from desktop to NVIDIA DGX™ cloud or other NVIDIA® accelerated data centers or cloud infrastructures, just with minimal code changes.
Designed for AI Developers
Data scientist / AI developer
AI developer who can fine tune AI model or AI inference application in local, easily migrate to company data center or cloud.
Academic
Practice AI training or develop AI inference application in classroom/lab. Migrate to school server room or AI laboratory to scale up.
Independent AI developer
Experiment with fine tuning AI models or run local inference to test or create local AI assistants. Easily migrate to cloud for more computing power.
Applications supported by NVIDIA® AI Frameworks
EdgeXpert provides an excellent platform for developing edge applications with NVIDIA® AI frameworks, including Isaac, Holoscan and many others.
Medical
Robot
Retail
Smart City
| EdgeXpert AI Supercomputer Specifications | |
|---|---|
| Architecture | NVIDIA® Grace Blackwell GB10 |
| GPU | NVIDIA® Blackwell Architecture |
| CPU | 20 core Arm (10 Cortex-X925 + 10 Cortex-A725) |
| Tensor Performance | 1000 AI TOPS (FP4, Sparse) |
| System Memory | 128 GB LPDDR5x (Unified) |
| Memory Interface | 256-bit |
| Memory Bandwidth | 273 GB/s |
| Storage | 1 or 4 TB NVMe M.2 (Self-Encrypting) |
| USB | 4x USB 3.2 Type-C |
| Ethernet | 1x RJ-45 (10 GbE) |
| NIC | ConnectX-7 Smart NIC |
| Wi-Fi / Bluetooth | Wi-Fi 7 / BT 5.4 |
| Display Connectors | 1x HDMI 2.1a, 3x DP1.4a via USB-C |
| OS | NVIDIA DGX™ OS |
| Dimensions | 151 × 151 × 52 mm (1.19L) |
| Weight | 1.2 kg |
Get Started with Notebook Plaza
Need a custom quote or want to speak to a rep? We're here to help.
HP ZGX Nano
A compact local AI station based on the NVIDIA® GB10 Grace Blackwell Superchip for building, fine-tuning and validating AI models.
Big AI Performance.
Tiny Footprint.
Prototype, fine-tune, and inference models of up to 200B parameters locally on an incredibly compact desktop.
NVIDIA® GB10 Grace Blackwell
Superchip
Get up to 1000 TOPS of AI compute with an NVIDIA® Blackwell GPU. Supercharge data preprocessing and orchestration with a Grace 20-core Arm CPU2.
Unified System Memory
Run AI development and testing workloads with AI models of up to 200 billion parameters at your desk with 128 GB of coherent unified system memory.
Fast, Secure Storage
Choose between 2 or 4 TB of NVMe M.2 self-encrypted storage3 to work efficiently with large files and keep more of your data local and secure.
Open Flexible Tools. Fast Setup
Use the built-in NVIDIA AI software stack and streamline device discovery, pairing, and installation of open-source tools with the HP ZGX Toolkit.1
Featuring: HP ZGX Toolkit
HP ZGX Toolkit:
Accelerate Time
to Results
Move from idea to deployment faster. Install the HP ZGX Toolkit1 and get open-source tools, MLflow, and Ollama. With instant discovery, sync, and export, train locally, track results, and deploy anywhere.
Learn More →
NVIDIA® ConnectX™
Networking
Work with even larger AI models locally—up to 405 billion parameters—by connecting two HP ZGX Nano systems together to scale local compute resources.4
Tiny AI Powerhouse
Measuring at 150mm L x 150mm W x 51mm H,5 this new class of desktop is purpose-built for AI development yet fits in the palm of your hand.
Offload AI
Workloads
Without the
Cloud
Pair existing systems—Windows, Mac, or Linux—with a network connected ZGX Nano. Get developer-grade performance without costly cloud instances and keep sensitive information local.
Advance Edge AI
Solutions
From real-time computer vision to running agentic and domain-specific models locally, get the performance needed to deploy edge AI solutions directly where data is generated.
HP ZGX Toolkit:
45%6 Faster Time to Results
From setup to starting your first workflow.
HP ZGX Toolkit vs “Do it yourself”
From 50 min to 30 min.
From 45 min to 25 min.
From 43 min to 23 min.
ZGX Nano for Secure &
Regulated Environments:
Get the ZGX Nano for Secure & Regulated Environments without Bluetooth or WiFi radios.
Same powerful compute, designed for environments with heightened security requirements.
Keep Sensitive
Data on Prem
Develop AI models on sensitive data without cloud transfer—fully air-gapped and secure. With the ZGX Nano for Secure & Regulated Environments, your data stays in the building.
Eliminate Cloud
Provisioning Friction
Skip approval queues, capacity waits, and tracking overhead. Developers can iterate instantly on dedicated hardware.
Built to Work
With the Cloud
Build and validate models locally on the HP ZGX Nano, then deploy confidently to your existing production infrastructure when ready. Move seamlessly from local to cloud and back for efficient AI development.
Contact your sales representative to learn more.
HP ZGX Nano Specifications
| Processor platform | NVIDIA® GB10 Grace Blackwell Superchip with a 20-core Arm CPU |
|---|---|
| GPU architecture | NVIDIA® Blackwell |
| AI performance | Up to 1000 AI TOPS at FP4 precision |
| Unified memory | 128 GB LPDDR5x coherent unified system memory |
| Storage | 2 TB or 4 TB NVMe M.2 self-encrypting storage |
| Networking | 10Gb Ethernet and NVIDIA ConnectX networking; two QSFP 200Gbps interfaces |
| Display / USB | HDMI 2.1a and three USB Type-C® 20Gbps ports, plus a USB Type-C® power connector |
| Wireless | Wi-Fi and Bluetooth on standard configurations; a radio-free configuration is available for secure environments |
| Operating system | NVIDIA DGX™ OS |
| Dimensions | Approximately 150 × 150 × 51 mm, excluding feet |
Get Started with Notebook Plaza
Need a custom quote or want to speak to a representative? Contact our sales team.
Dell Pro Max with GB10
A compact AI accelerator built for local inferencing, AI development and versatile edge workloads.
Discover Dell Pro Max
with GB10
Massive output. Small footprint.
Delivering exceptional performance in a compact design, this system features the cutting-edge NVIDIA® Grace-Blackwell architecture. With NVLink-C2C for ultra-low latency and 128GB LPDDR5X unified memory, it streamlines data sharing between CPU and GPU, enhancing efficiency and supporting demanding AI models like GenAI and LLMs.
Powerful, localised AI solutions
Take on AI and data science workloads with ease, powered by enabled large model local inferencing—up to 200B parameters.
Scale
Connect two Dell Pro Max with GB10 utilising the ConnectX-7 port to work with larger models and workloads.
Build
NVIDIA AI software stack provides developers with tools to build and run AI workflows.
Secure
Data stays local without having to transfer to external locations.
Control
Your own personal AI cloud—no contention for data centre compute or expense of cloud instances.
Scaled for success
Meet next-level efficiency using scalable inferencing with faster performance compared to traditional multi-GPU setups. Equipped with the ConnectX-7 SmartNIC and 200GbE connectivity, this system enables seamless high-speed clustering across nodes to run large models efficiently. With Remote Direct Memory Access (RDMA) and GPUDirect, CPUs and GPUs can access each other’s memory directly, eliminating overhead and enabling secure, shared workloads.
Seamless performance
Stable, fully-tested environments efficiently run AI, machine learning, and analytics workloads ensuring reliability and maximum productivity for advanced applications. The NVIDIA DGX OS, built on Ubuntu Linux, is tailored with system-specific optimisations and additional drivers for simplified efficiency.
Innovate securely at the edge
The edge is where your data comes to life. With TPM 2.0, fast AI data collection and secure analytics on the edge you’ll get faster discoveries, streamlined workflows with the ability to innovate effectively. Working at the source empowers you to uncover hidden efficiencies, create better experiences, and deliver solutions that truly drive progress.
Ports and slots
- Power button
- USB-C Power-In
- Three USB-C Gen 2x2 ports with DisplayPort Alt Mode
- HDMI 2.1b
- Network (RJ-45, 10GbE)
- ConnectX-7 Smart NIC with two 200G QSFP interfaces; one is required to connect to another Dell Pro Max with GB10
Dimensions and weight
- 1. Height (max)2.00 in. (51.00 mm)
- 2. Depth5.90 in. (150.00 mm)
- 3. Width5.90 in. (150.00 mm)
Weight (minimum): 2.89 lb (1.31 kg)
Dell Pro Max with GB10 Specifications
| Processor | NVIDIA GB10 with 16 MB L2 cache and 20 Arm cores |
|---|---|
| Operating system | NVIDIA DGX™ OS |
| Graphics | Integrated NVIDIA Blackwell graphics |
| AI performance | Up to one petaflop / 1000 TFLOPS of FP4 compute |
| Memory | 128 GB LPDDR5x coherent unified system memory, 273 GB/s |
| Storage options | 2 TB M.2 2230 QLC SSD; or 4 TB M.2 2242 self-encrypting Opal 2.0 SSD |
| Network | RJ-45 10GbE and ConnectX-7 Smart NIC with two 200G QSFP interfaces |
| USB / display | USB-C power input; three USB-C Gen 2x2 ports with DisplayPort Alt Mode; HDMI 2.1b |
| Wireless | Wi-Fi 7 (802.11be), 2x2, Bluetooth® 5.4, internal antenna |
| Dimensions | 150 mm wide × 150 mm deep; 45.5 mm front/rear height and 51 mm peak height |
| Minimum weight | 1.31 kg |
| Chassis | Magnetite anodized aluminium main chassis with painted front/rear bezel and black elastomer foot |
| Power | 280 W USB Type-C AC adapter |
| Optical drive / slots | None |
| Support | Basic Hardware Service, ProSupport and ProSupport Plus options with onsite service after remote diagnosis |
Coming Soon
Product details for this AI Supercomputer brand will be available soon. Stay tuned!