HP ZGX Fury Now Available: GB300 Superchip, 748GB Unified Memory
HP launches its AI station with Red Hat and NVIDIA collaboration for edge computing.
1Introduction to HP ZGX Fury
HP has announced the availability of its ZGX Fury AI station, designed for edge computing. The system features NVIDIA's GB300 Grace Blackwell Ultra Desktop Superchip, offering 748GB of unified memory and up to 20 petaFLOPS of FP4 compute. HP positions this as a shared inference box for departments or factories, eliminating the need for a data center.
2Core Specifications
The ZGX Fury is built around NVIDIA's Blackwell Ultra GPU with 252GB of HBM3e memory and a 72-core Grace CPU. The system includes 748GB of coherent memory, allowing the GPU to directly access CPU memory. This setup enables trillion-parameter inference and fine-tuning of models with up to 100 billion parameters on a single machine.
3Storage and Networking
The ZGX Fury includes four M.2 slots: two for the operating system in RAID 1 and two for data storage in RAID 0. Networking features include ConnectX-8 with two QSFP112 ports at 400Gbps each, enabling high-speed connections between systems. Additional ports include USB-A, USB-C, audio jacks, and a Kensington slot.
4Cooling and Design
The chassis is a tower design with optional 5U rack rails. HP uses liquid cooling with optimized airflow, similar to the MSI XpertStation WS300. The cooling system maintains the GPU at 71°C under full load, ensuring efficient performance without overheating.
5Software and Tools
HP ships the ZGX Fury with Ubuntu 24.04 LTS and NVIDIA's AI developer tools. Additional software includes HP Z Runtime for model management and HP Z Toolkit for open-source frameworks and MLflow experiment tracking. These tools facilitate local model deployment and management.
6Red Hat Collaboration
HP has partnered with Red Hat to certify the ZGX Fury for Red Hat Enterprise Linux. The collaboration aims to develop an open, enterprise-grade AI platform that integrates Red Hat AI Factory with NVIDIA. This platform reduces setup time, improves GPU utilization, and supports multiple AI workloads with workload isolation.
7Market Positioning
HP positions the ZGX Fury as a solution for local AI deployment, extending enterprise AI from data centers to the edge. The system competes with MSI's WS300 and AMD's Threadripper Halo Station, targeting similar workloads. HP emphasizes greater choice, control, and consistency for customers deploying local AI factories.
8Future Implications
The ZGX Fury represents a shift towards localized AI processing, enabling powerful AI capabilities closer to where decisions are made. This development reduces reliance on centralized data centers, offering scalability and security for enterprise applications. HP's collaboration with Red Hat and NVIDIA sets a new standard for edge AI infrastructure.