HP ZGX Fury G1n GB300 AI Workstation with NVIDIA Blackwell Ultra

HP ZGX Fury G1n GB300 AI Workstation with NVIDIA Blackwell Ultra, 748GB Coherent Memory, 72-Core Grace CPU & Dual 400G ConnectX-8 Networking - Local Warranty 

Next-Generation Enterprise AI Workstation Powered by NVIDIA GB300

The HP ZGX Fury G1n GB300 AI Workstation is a next-generation AI supercomputer designed to accelerate the most demanding artificial intelligence, machine learning, and high-performance computing workloads. Built around the NVIDIA Grace Blackwell Ultra (GB300) Superchip, it delivers up to 20 PFLOPS of AI compute, empowering organisations to train, fine-tune, and deploy large language models (LLMs), AI agents, generative AI applications, computer vision models, and advanced scientific simulations entirely within their own infrastructure. It is purpose-built for enterprises, research institutions, universities, government agencies, and AI innovation labs seeking maximum performance with complete data sovereignty.

748GB Coherent Unified Memory for Large-Scale AI Models

The HP ZGX Fury G1n features an impressive 748GB of coherent unified memory, enabling AI models and massive datasets to reside in a single memory space for significantly improved performance and efficiency. This unified architecture reduces data movement between CPU and GPU, accelerating AI inferencing, retrieval-augmented generation (RAG), foundation model fine-tuning, engineering simulations, digital twins, cybersecurity analytics, and complex research workloads.

72-Core NVIDIA Grace CPU with Dual 400G ConnectX-8 Networking

At the heart of the workstation is the 72-core NVIDIA Grace Arm Neoverse V2 CPU, engineered to work seamlessly with the Blackwell Ultra GPU through NVIDIA's high-bandwidth coherent architecture. Combined with dual NVIDIA ConnectX®-8 400GbE QSFP networking, the system delivers ultra-low latency communication and exceptional bandwidth for distributed AI training, private AI clusters, and high-performance computing environments, ensuring outstanding scalability for enterprise deployments.

Enterprise-Ready AI Infrastructure for Mission-Critical Workloads

Designed for continuous operation, the HP ZGX Fury G1n offers enterprise-class reliability, advanced thermal engineering, secure remote management, and support for NVIDIA Multi-Instance GPU (MIG) technology to maximise resource utilisation across multiple users and AI workloads. Whether deployed as a standalone AI workstation or integrated into a private AI cloud, it provides the performance, security, and scalability required for healthcare, finance, manufacturing, engineering, autonomous systems, and government AI initiatives.

Technical Specifications

General

Model: HP ZGX Fury G1n
Product Name: HP ZGX Fury G1n GB300 AI Workstation with NVIDIA Blackwell Ultra, 748GB Coherent Memory, 72-Core Grace CPU & Dual 400G ConnectX-8 Networking
Manufacturer: HP
Product Family: HP ZGX
Product Series: HP ZGX Fury
Product Type: Enterprise AI Workstation / Deskside AI Supercomputer
Compute Platform: NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
Processor Architecture: NVIDIA Grace CPU + NVIDIA Blackwell Ultra GPU
CPU Platform: NVIDIA Grace CPU
GPU Platform: NVIDIA Blackwell Ultra
System Architecture: Coherent CPU-GPU Accelerated Computing Platform
CPU Core Count: 72 Arm Neoverse V2 Cores
Total Coherent Memory: 748GB
AI Compute Performance: Up to 20 PetaFLOPS FP4
Supported AI Model Size: Up to 1 Trillion Parameters
High-Speed Networking: Dual 400G NVIDIA ConnectX-8 Networking
Maximum Aggregate High-Speed Network Bandwidth: Up to 800Gbps
Deployment Type: Deskside, Enterprise Laboratory, AI Development Facility or Rack-Ready Environment
Primary Workloads: Generative AI, Agentic AI, Large Language Models, Model Training, Fine-Tuning, Inferencing, Data Science and High-Performance Computing
Target Users: Enterprises, Government Agencies, Universities, Research Institutions, AI Developers, Data Scientists and Engineering Teams
Operating System: NVIDIA AI-Optimised Linux Environment, Configuration Dependent
NVIDIA AI Software Stack: Supported
Enterprise Remote Management: Supported, Configuration Dependent
Warranty: HP Enterprise Workstation Warranty, Configuration Dependent

Processor Specifications

Processor Platform: NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
CPU Manufacturer: NVIDIA
CPU Model: NVIDIA Grace CPU
CPU Quantity: 1
CPU Architecture: Arm
CPU Core Architecture: Arm Neoverse V2
CPU Core Count: 72 Cores
CPU Class: Datacentre-Class Accelerated Computing Processor
CPU-GPU Interconnect: NVIDIA NVLink-C2C
CPU-GPU Coherency: Supported
Unified CPU-GPU Address Space: Supported
High-Bandwidth CPU-GPU Communication: Supported
Low-Latency Data Exchange: Supported
Large-Scale Parallel Computing: Supported
Multi-Threaded AI Processing: Supported
Scientific Computing: Supported
High-Performance Computing: Supported
Data Science Processing: Supported
Continuous Enterprise Workloads: Supported
AI Pipeline Processing: Supported
Enterprise Model Development: Supported
Optimised for NVIDIA CUDA-X Libraries
Optimised for NVIDIA AI Enterprise Software
Optimised for Large Language Model Workloads

AI Performance

Maximum AI Compute Performance: Up to 20 PetaFLOPS FP4
AI Accelerator: NVIDIA Blackwell Ultra GPU
GPU Generation: NVIDIA Blackwell Ultra
Tensor Core Generation: Fifth-Generation NVIDIA Tensor Cores
Transformer Engine: Second-Generation NVIDIA Transformer Engine
FP4 AI Computing: Supported
FP6 AI Computing: Supported
FP8 AI Computing: Supported
FP16 AI Computing: Supported
BF16 AI Computing: Supported
TF32 Computing: Supported
INT8 AI Inferencing: Supported
Mixed-Precision Computing: Supported
Microscaling Data Formats: Supported
Dynamic Precision Management: Supported
AI Reasoning Acceleration: Supported
Generative AI Acceleration: Supported
Large Language Model Acceleration: Supported
Foundation Model Training: Supported
Foundation Model Fine-Tuning: Supported
Production AI Inferencing: Supported
Long-Context Inferencing: Supported
Multimodal AI Processing: Supported
AI Agent Development: Supported
Long-Running AI Agent Support: Supported
Retrieval-Augmented Generation: Supported
Scientific AI Processing: Supported
High-Performance Data Analytics: Supported
Supported AI Model Size: Up to 1 Trillion Parameters

AI Features

Large Language Model Development
Large Language Model Training
Large Language Model Fine-Tuning
Large Language Model Inferencing
Foundation Model Development
Foundation Model Optimisation
Generative AI Application Development
Agentic AI Development
Autonomous AI Agents
Long-Running AI Agents
Reasoning Model Processing
Retrieval-Augmented Generation
Multimodal AI Processing
Natural Language Processing
Computer Vision
Speech Recognition
Speech Synthesis
Recommendation Systems
Predictive Analytics
Digital Twin Simulation
Scientific Machine Learning
Physics-Informed Machine Learning
Drug Discovery
Molecular Modelling
Genomics Research
Medical Imaging Analysis
Financial Modelling
Fraud Detection
Cybersecurity AI
Manufacturing AI
Autonomous Systems Development
Robotics AI
Enterprise Knowledge Assistants
Private AI Deployment
Sovereign AI Deployment
On-Premise Model Development
Local AI Inferencing
Hybrid Cloud AI Development
Datacentre AI Deployment
Production Model Serving
Multi-User AI Development
Enterprise MLOps
AI Factory Development

Memory Specifications

Total System Memory: 748GB
Memory Architecture: NVIDIA Coherent Unified Memory
CPU-GPU Shared Memory Pool: Supported
Unified Address Space: Supported
Hardware Memory Coherency: Supported
Direct CPU and GPU Memory Access: Supported
High-Bandwidth Memory Architecture: Supported
Low-Latency Memory Access: Supported
Reduced CPU-GPU Data Copying: Supported
Large AI Model Loading: Supported
Large Context Window Processing: Supported
Trillion-Parameter Model Inferencing: Supported
Multi-Billion Parameter Model Fine-Tuning: Supported
Large Dataset Processing: Supported
In-Memory Data Science: Supported
AI Model Checkpoint Processing: Supported
Retrieval-Augmented Generation Support: Supported
Vector Database Processing: Supported
Large Embedding Database Processing: Supported
High-Concurrency Inferencing: Supported
Multi-User AI Workloads: Supported
Scientific Simulation Support: Supported
High-Performance Data Analytics: Supported
Memory-Intensive AI Pipeline Support: Supported
Enterprise AI Application Support: Supported

GPU Specifications

GPU Manufacturer: NVIDIA
GPU Architecture: NVIDIA Blackwell Ultra
GPU Type: Datacentre-Class Enterprise AI Accelerator
GPU Integration: Integrated with NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
Tensor Cores: Fifth-Generation NVIDIA Tensor Cores
Transformer Engine: Second-Generation NVIDIA Transformer Engine
NVIDIA CUDA Support: Supported
NVIDIA CUDA-X Support: Supported
NVIDIA cuDNN Support: Supported
NVIDIA TensorRT Support: Supported
NVIDIA TensorRT-LLM Support: Supported
NVIDIA Triton Inference Server Support: Supported
NVIDIA RAPIDS Support: Supported
NVIDIA NeMo Support: Supported
NVIDIA NGC Container Support: Supported
NVIDIA AI Enterprise Support: Supported
NVIDIA GPUDirect RDMA: Supported
NVIDIA GPUDirect Storage: Supported
Mixed-Precision AI Computing: Supported
FP4 Tensor Processing: Supported
FP8 Tensor Processing: Supported
BF16 Tensor Processing: Supported
AI Training Acceleration: Supported
AI Fine-Tuning Acceleration: Supported
AI Inferencing Acceleration: Supported
Generative AI Acceleration: Supported
AI Reasoning Acceleration: Supported
Scientific Simulation Acceleration: Supported
Data Analytics Acceleration: Supported
Visualisation Support: Configuration Dependent
Additional NVIDIA RTX PRO GPU Support: Configuration Dependent

Storage Specifications

Storage Technology: Enterprise NVMe Solid-State Drive
Storage Interface: PCI Express NVMe
Storage Generation: Configuration Dependent
Installed Storage Capacity: Configuration Dependent
Primary Boot Storage: NVMe SSD
Enterprise NVMe SSD Support: Supported
Storage Expansion: Configuration Dependent
M.2 NVMe Storage Support: Configuration Dependent
PCIe Storage Expansion: Supported, Configuration Dependent
RAID Support: Configuration Dependent
Storage Redundancy: Configuration Dependent
Self-Encrypting Drive Support: Drive and Configuration Dependent
High-Speed AI Dataset Loading: Supported
Large AI Model Repository Support: Supported
AI Model Checkpoint Storage: Supported
Vector Database Storage: Supported
Training Dataset Storage: Supported
Container Image Storage: Supported
Local Model Cache Support: Supported
High-Speed Data Preprocessing: Supported
Continuous Enterprise Read and Write Workloads: Supported
Storage Health Monitoring: Configuration Dependent
Remote Storage Status Monitoring: Configuration Dependent

Networking Specifications

High-Speed Network Platform: NVIDIA ConnectX-8
High-Speed Network Controller Quantity: 2
High-Speed Network Ports: 2 × 400G
Maximum Per-Port Network Bandwidth: Up to 400Gbps
Maximum Aggregate Network Bandwidth: Up to 800Gbps
Network Connector Type: Configuration Dependent
NVIDIA ConnectX-8 SuperNIC Support: Supported
High-Speed Ethernet Support: Supported
InfiniBand Support: Configuration Dependent
RDMA Support: Supported
RoCE Support: Supported
RoCE v2 Support: Supported
NVIDIA GPUDirect RDMA: Supported
NVIDIA GPUDirect Storage: Supported
Ultra-Low-Latency Networking: Supported
High-Throughput Networking: Supported
Distributed AI Training: Supported
Multi-Node Model Training: Supported
AI Cluster Interconnect: Supported
High-Performance Storage Fabric: Supported
Scale-Out AI Infrastructure: Supported
Enterprise AI Fabric Integration: Supported
High-Performance Computing Cluster Support: Supported
Rapid Model Distribution: Supported
Large Dataset Transfer: Supported
Datacentre Network Integration: Supported

Ports & Expansion

2 × 400G NVIDIA ConnectX-8 High-Speed Network Connections
Enterprise PCI Express Expansion: Supported
PCIe Expansion Generation: Configuration Dependent
Additional NVIDIA RTX PRO GPU Expansion: Configuration Dependent
High-Speed Storage Expansion: Supported
Enterprise Network Adapter Expansion: Supported
Storage Controller Expansion: Supported
Specialised Accelerator Expansion: Configuration Dependent
USB Connectivity: Configuration Dependent
Display Connectivity: Configuration Dependent
Dedicated Management Interface: Configuration Dependent
Audio Connectivity: Configuration Dependent
Rack Integration: Supported, Configuration Dependent
Future Enterprise Hardware Expansion: Supported

Additional GPU Expansion

Additional GPU Support: Up to 1 × NVIDIA RTX PRO Blackwell Generation GPU, Configuration Dependent
Professional Visualisation Acceleration: Supported with Optional GPU
Ray-Traced Rendering: Supported with Compatible Optional GPU
AI-Assisted Rendering: Supported
Engineering Simulation: Supported
Digital Content Creation: Supported
Computer-Aided Design: Supported
Computer-Aided Engineering: Supported
Visual Computing: Supported
Synthetic Data Generation: Supported
Omniverse Workloads: Supported
Digital Twin Visualisation: Supported
AI and Graphics Converged Workflows: Supported
Optional GPU Availability: Configuration Dependent
Optional GPU Power Requirement: Configuration Dependent

NVIDIA NVLink-C2C Architecture

CPU-GPU Interconnect: NVIDIA NVLink-C2C
Coherent Interconnect: Supported
Unified Memory Access: Supported
High-Bandwidth Communication: Supported
Low-Latency CPU-GPU Data Transfer: Supported
Shared Virtual Addressing: Supported
Hardware Cache Coherency: Supported
Reduced Data Movement: Supported
Improved AI Pipeline Efficiency: Supported
Large Model Processing: Supported
Memory-Intensive Data Science: Supported
Scientific Computing Acceleration: Supported
Integrated Superchip Communication: Supported

Operating System

Operating System Platform: NVIDIA AI-Optimised Linux
Operating System Version: Configuration Dependent
Ubuntu Linux Compatibility: Supported
NVIDIA GPU Driver Stack: Supported
NVIDIA CUDA Development Environment: Supported
NVIDIA Container Toolkit: Supported
Docker Support: Supported
Kubernetes Support: Supported
Python AI Development: Supported
C++ CUDA Development: Supported
AI Framework Compatibility: Supported
Remote Administration: Supported
Command-Line Administration: Supported
Enterprise Package Management: Supported
Security Update Support: Supported
Containerised Workload Support: Supported
Cloud-Native AI Development: Supported
Datacentre Deployment Compatibility: Supported

NVIDIA AI Software Support

NVIDIA AI Enterprise
NVIDIA CUDA Toolkit
NVIDIA CUDA-X Libraries
NVIDIA cuDNN
NVIDIA TensorRT
NVIDIA TensorRT-LLM
NVIDIA Triton Inference Server
NVIDIA NeMo
NVIDIA NeMo Framework
NVIDIA NeMo Retriever
NVIDIA RAPIDS
NVIDIA NGC Containers
NVIDIA Container Toolkit
NVIDIA NCCL
NVIDIA Nsight Systems
NVIDIA Nsight Compute
NVIDIA Base Command Manager Compatibility
NVIDIA Omniverse Compatibility
NVIDIA NIM Microservices
NVIDIA Blueprints
NVIDIA AI Workbench
NVIDIA DGX Cloud Integration
NVIDIA Enterprise Support, Licence Dependent

AI Framework Support

PyTorch
TensorFlow
JAX
ONNX Runtime
Hugging Face Transformers
DeepSpeed
Megatron-LM
vLLM
TensorRT-LLM
NVIDIA NeMo
NVIDIA Triton Inference Server
NVIDIA RAPIDS
Ray
Apache Spark
MLflow
Kubeflow
Docker
Kubernetes
Red Hat OpenShift
Vector Database Platforms
Enterprise MLOps Platforms
Distributed Training Frameworks
AI Model Serving Frameworks
Scientific Computing Libraries
Python Development Frameworks
CUDA C++ Development Tools

Container & Orchestration Support

Docker Containers
NVIDIA Container Toolkit
NVIDIA NGC Containers
Kubernetes
Red Hat OpenShift
Containerised AI Development
Containerised Model Training
Containerised Model Inferencing
GPU-Accelerated Containers
Cloud-Native AI Deployment
Enterprise MLOps Integration
Workload Scheduling
Distributed Training Orchestration
Model Serving Orchestration
Multi-User Development Environments
Reproducible AI Environments
Private Cloud Integration
Hybrid Cloud Integration
Datacentre Deployment Integration

Security Features

HP Enterprise Workstation Security
Secure Boot Support
Hardware Root of Trust: Configuration Dependent
Trusted Platform Module 2.0: Configuration Dependent
Firmware Integrity Protection
BIOS Protection
Secure Firmware Update Support
Hardware-Based Platform Validation
Platform Integrity Verification
Secure Operating System Environment
Role-Based Administrative Access
Secure Remote Management
Encrypted Management Sessions
Management Network Isolation: Configuration Dependent
System Event Logging
Audit Logging
Data-at-Rest Encryption: Drive and Configuration Dependent
Self-Encrypting Drive Support: Configuration Dependent
Enterprise Authentication Integration
Physical Security Features: Configuration Dependent
Secure Model and Dataset Processing
On-Premise Data Control
Private AI Workload Support

HP Enterprise Security Features

HP Sure Start: Configuration Dependent
HP Sure Admin: Configuration Dependent
HP Sure Run: Configuration Dependent
HP Sure Recover: Configuration Dependent
HP Sure Click: Operating System and Configuration Dependent
HP BIOSphere: Configuration Dependent
HP Secure Erase: Supported on Compatible Storage
HP Tamper Lock: Configuration Dependent
HP Client Security Manager: Configuration Dependent
Platform Certificate Support: Configuration Dependent
Enterprise Device Identity: Configuration Dependent
Centralised Security Management: Configuration Dependent
Security Feature Availability: Dependent on Final HP Configuration

Management Features

Enterprise Remote Management: Supported
Out-of-Band Management: Configuration Dependent
Dedicated Management Controller: Configuration Dependent
Remote Console Access: Configuration Dependent
Remote Power Control: Configuration Dependent
Remote System Reset: Configuration Dependent
Remote Firmware Updates: Supported
Remote BIOS Configuration: Supported
System Health Monitoring: Supported
Processor Monitoring: Supported
GPU Monitoring: Supported
Memory Monitoring: Supported
Storage Monitoring: Supported
Network Monitoring: Supported
Temperature Monitoring: Supported
Fan Monitoring: Supported
Voltage Monitoring: Supported
Power Monitoring: Supported
Hardware Inventory: Supported
Asset Management: Supported
System Event Logging: Supported
Alert and Notification Support: Supported
Enterprise Fleet Management: Supported
HP Management Software Compatibility: Configuration Dependent
Datacentre Management Integration: Supported

Virtualisation Features

Container-Based Virtualisation: Supported
Virtual Machine Support: Configuration Dependent
GPU Virtualisation: Configuration Dependent
Multi-User AI Development: Supported
Workload Isolation: Supported
Containerised Development Environments: Supported
Cloud-Native AI Workloads: Supported
Private Cloud Integration: Supported
Hybrid Cloud Integration: Supported
Enterprise Resource Scheduling: Supported
AI Workload Orchestration: Supported
Development Sandbox Support: Supported
Department-Level Resource Allocation: Supported
Enterprise MLOps Integration: Supported

Data Science Features

In-Memory Data Processing
Large Dataset Ingestion
GPU-Accelerated Data Preparation
GPU-Accelerated Data Analytics
NVIDIA RAPIDS Support
Apache Spark Integration
Dataframe Acceleration
Machine Learning Model Development
Statistical Analysis
Feature Engineering
Model Validation
Model Optimisation
Experiment Tracking
Interactive Data Science
Large Data Lake Processing
Local Dataset Processing
Reduced Cloud Data Transfer Requirements
Private Data Analysis
Sensitive Dataset Processing
Scientific Visualisation, Configuration Dependent

Power Specifications

Power Supply Type: HP Enterprise High-Efficiency Power Supply
Power Supply Capacity: Configuration Dependent
Power Input: Configuration Dependent
Power Consumption: Workload and Configuration Dependent
High-Load AI Workload Support: Supported
Continuous GPU Workload Support: Supported
Remote Power Monitoring: Configuration Dependent
Remote Power Control: Configuration Dependent
Power Usage Reporting: Configuration Dependent
Power Fault Monitoring: Supported
Enterprise Power Management: Supported
Datacentre Power Integration: Supported
Energy-Efficient AI Processing: Supported
Power Supply Redundancy: Configuration Dependent
Power Supply Efficiency Certification: Configuration Dependent

Cooling & Reliability

Professional Enterprise Workstation Cooling
High-Efficiency Thermal Management
Datacentre-Class Component Cooling
CPU Thermal Monitoring
GPU Thermal Monitoring
Memory Thermal Monitoring
Storage Thermal Monitoring
Variable-Speed Fan Control
Automated Fan Management
Thermal Protection
Hardware Health Alerts
Long-Duration AI Training Support
Continuous AI Inferencing Support
24×7 Operational Design
Enterprise-Grade Components
High-Bandwidth Component Cooling
Optional GPU Cooling Support
High-Speed Networking Cooling Support
Serviceable Thermal Components
Acoustic Performance: Configuration Dependent
Dedicated Facility Cooling: Not Required for Standard Deskside Deployment

Compatibility

Compatible AI Workloads:

Large Language Models
Trillion-Parameter Model Inferencing
Foundation Models
Generative AI
AI Agents
Agentic AI
Reasoning Models
Retrieval-Augmented Generation
Multimodal AI
Computer Vision
Natural Language Processing
Speech AI
Recommendation Engines
Scientific Machine Learning
Predictive Analytics
Digital Twins
Autonomous Systems
Robotics AI
Enterprise AI Inferencing
Model Training
Model Fine-Tuning
Model Optimisation
Model Serving
Synthetic Data Generation
High-Performance Data Analytics

Compatible Development Frameworks:

PyTorch
TensorFlow
JAX
ONNX Runtime
Hugging Face Transformers
DeepSpeed
Megatron-LM
vLLM
NVIDIA NeMo
NVIDIA TensorRT
NVIDIA TensorRT-LLM
NVIDIA Triton Inference Server
NVIDIA RAPIDS

Compatible Deployment Platforms:

NVIDIA AI Enterprise
NVIDIA NGC
Docker
Kubernetes
Red Hat OpenShift
Enterprise MLOps Platforms
Private Cloud Infrastructure
Hybrid Cloud Infrastructure
AI Factory Infrastructure
Scientific Computing Environments
Enterprise DevOps Environments
Datacentre AI Infrastructure

Compatible Network Environments:

400G High-Speed Networks
NVIDIA ConnectX-8 Networking
RDMA Networks
RoCE Networks
High-Performance Storage Networks
Distributed AI Clusters
High-Performance Computing Clusters
Enterprise Ethernet Fabrics
AI Factory Network Fabrics

Design & Construction

Professional Enterprise AI Workstation Design
Deskside AI Supercomputer Architecture
Rack-Ready Deployment: Configuration Dependent
High-Density Compute Platform
Datacentre-Class Internal Architecture
Enterprise Cable Management
High-Speed Expansion Architecture
Serviceable Internal Components
Continuous 24×7 Operational Design
Remote Serviceability
Enterprise Laboratory Deployment Ready
Datacentre Integration Ready
Office and Engineering Workspace Ready
AI Research Environment Ready
Professional HP Z Workstation Engineering
Tool-Less Service Access: Configuration Dependent
Chassis Security: Configuration Dependent

Environmental Specifications

Operating Environment: Enterprise Office, Research Laboratory or Datacentre
Operating Temperature: Configuration Dependent
Storage Temperature: Configuration Dependent
Operating Humidity: Configuration Dependent
Storage Humidity: Configuration Dependent
Maximum Operating Altitude: Configuration Dependent
Thermal Monitoring: Supported
Environmental Monitoring: Supported
Continuous Operation: Supported
ENERGY STAR Certification: Configuration Dependent
EPEAT Registration: Region and Configuration Dependent
TCO Certification: Configuration Dependent
RoHS Compliance: Region Dependent
Low-Halogen Materials: Configuration Dependent
Sustainable Material Content: Configuration Dependent
Compliance and Certifications: Region and Configuration Dependent

Physical Specifications

Product Type: Enterprise AI Workstation
Form Factor: Deskside AI Supercomputer
Rack-Ready Capability: Configuration Dependent
Chassis Colour: Black, Configuration Dependent
Dimensions: Configuration Dependent
System Weight: Configuration Dependent
Installed Weight: Configuration Dependent
Service Access: Enterprise Workstation Service Design
Internal Expansion Access: Supported
Optional GPU Access: Supported
High-Speed Network Access: Dual 400G Connections
Rack Mounting Kit: Configuration Dependent
Cable Management Accessories: Configuration Dependent

Package Contents

HP ZGX Fury G1n AI Workstation
NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
NVIDIA Grace 72-Core CPU
NVIDIA Blackwell Ultra GPU
748GB Coherent Unified Memory
Dual NVIDIA ConnectX-8 400G Networking
Pre-Installed Operating System, Configuration Dependent
NVIDIA AI Software Environment, Configuration Dependent
HP Keyboard: Configuration Dependent
HP Mouse: Configuration Dependent
Power Cable
Quick Installation Guide
Safety Documentation
Warranty Documentation
HP Enterprise Support Information

Key Features

NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
Single NVIDIA Grace CPU
72 Arm Neoverse V2 CPU Cores
Single NVIDIA Blackwell Ultra GPU
Fifth-Generation NVIDIA Tensor Cores
Second-Generation NVIDIA Transformer Engine
Up to 20 PetaFLOPS FP4 AI Performance
748GB Coherent Unified Memory
Unified CPU-GPU Address Space
NVIDIA NVLink-C2C Interconnect
Support for AI Models Up to 1 Trillion Parameters
Dual 400G NVIDIA ConnectX-8 Networking
Up to 800Gbps Aggregate High-Speed Network Bandwidth
NVIDIA GPUDirect RDMA Support
NVIDIA GPUDirect Storage Support
Optional NVIDIA RTX PRO Blackwell GPU Support
NVIDIA AI Enterprise Ready
NVIDIA CUDA-X Ready
NVIDIA NGC Container Support
Large Language Model Training and Inferencing
Generative AI Development
Agentic AI Development
AI Reasoning Workloads
Enterprise Data Science
High-Performance Computing
Deskside AI Supercomputer Design
Rack-Ready Deployment, Configuration Dependent
Private and On-Premise AI Development
Enterprise HP Security and Management

Performance & Features

The HP ZGX Fury G1n is an enterprise AI workstation powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. It integrates a 72-core NVIDIA Grace CPU based on the Arm Neoverse V2 architecture with an NVIDIA Blackwell Ultra GPU, creating a tightly coupled accelerated computing platform for large-scale artificial intelligence, data science and high-performance computing workloads.

The system delivers up to 20 petaFLOPS of FP4 AI compute performance through fifth-generation NVIDIA Tensor Cores and the second-generation NVIDIA Transformer Engine. Support for advanced low-precision data formats improves throughput for generative AI, agentic AI, reasoning models, large language model inferencing and other compute-intensive AI workloads.

A total of 748GB coherent unified memory allows the NVIDIA Grace CPU and Blackwell Ultra GPU to access a shared memory pool through the NVIDIA NVLink-C2C interconnect. This architecture reduces unnecessary data movement between separate CPU and GPU memory spaces, improves processing efficiency and enables substantially larger AI models and datasets to run locally.

The large coherent memory pool enables the HP ZGX Fury G1n to support AI models containing up to one trillion parameters. This makes the workstation suitable for developing, fine-tuning and inferencing large language models, multimodal models, foundation models and autonomous AI agents without relying entirely on remote cloud infrastructure.

Dual NVIDIA ConnectX-8 400G network interfaces provide up to 800Gbps of aggregate high-speed connectivity. Support for RDMA, RoCE, NVIDIA GPUDirect RDMA and GPUDirect Storage enables low-latency communication between compute systems, storage platforms and GPU resources in distributed AI and high-performance computing environments.

The high-speed networking architecture allows the workstation to operate as an individual deskside AI system or as part of a larger AI development cluster. Organisations can use multiple systems for distributed model training, high-throughput dataset processing, model serving and collaborative AI research.

Optional support for an additional NVIDIA RTX PRO Blackwell Generation GPU allows organisations to combine datacentre-class AI computing with professional graphics, ray-traced rendering, simulation, digital twins and advanced visualisation. Final optional GPU support depends on the selected HP system configuration.

The workstation supports the NVIDIA AI software platform, including CUDA, CUDA-X libraries, TensorRT, TensorRT-LLM, Triton Inference Server, NeMo, RAPIDS, NGC containers and NVIDIA AI Enterprise. It is compatible with commonly used frameworks such as PyTorch, TensorFlow, JAX, ONNX Runtime, Hugging Face Transformers and DeepSpeed.

Its deskside form factor provides datacentre-class AI capabilities without requiring a specialised liquid-cooled rack installation. This allows AI development teams, researchers and data scientists to build and test large models locally while maintaining direct control over sensitive datasets, intellectual property and model assets.

The HP ZGX Fury G1n is designed to support a local-to-datacentre-to-cloud AI workflow. Models and applications can be developed and validated locally before being deployed to compatible NVIDIA-accelerated datacentre infrastructure or cloud platforms.

Typical Use Cases

Trillion-Parameter Model Inferencing
Large Language Model Training
Large Language Model Fine-Tuning
Large Language Model Optimisation
Foundation Model Development
Generative AI Application Development
AI Agent Development
Agentic AI Deployment
Autonomous Long-Running AI Agents
AI Reasoning Model Processing
Retrieval-Augmented Generation
Multimodal AI
Production AI Inferencing
Machine Learning
Deep Learning
Computer Vision
Natural Language Processing
Speech Recognition
Speech Generation
Recommendation Systems
Enterprise Knowledge Assistants
Cybersecurity AI
Financial Services AI
Fraud Detection
Healthcare AI Research
Medical Imaging Analysis
Drug Discovery
Molecular Simulation
Genomics Research
Scientific Computing
High-Performance Computing
Digital Twin Simulation
Manufacturing AI
Autonomous Systems Development
Robotics Research
Synthetic Data Generation
GPU-Accelerated Data Science
Large Dataset Analytics
University Research Laboratories
Government AI Infrastructure
Private AI Development
Sovereign AI Infrastructure
On-Premise AI Deployment
Hybrid Cloud AI Development
Enterprise MLOps
Multi-User AI Development
Distributed AI Training
High-Speed AI Cluster Nodes
Enterprise AI Factories
Professional Visualisation
Engineering Simulation
AI-Assisted Content Creation