HP unveils ZGX Fury AI Station for advanced computing

HP has announced the launch of the ZGX Fury AI Station, developed in collaboration with Red Hat, aimed at enterprises and advanced work environments. This workstation is designed to handle heavy artificial intelligence workloads, including model training and autonomous agent management without a constant reliance on the cloud. The system utilizes the Nvidia GB300 Grace Blackwell Ultra chip, featuring a powerful central processing unit with 72 ARM Neoverse V2 architecture cores and a massive 496 gigabytes of LPDDR5X RAM.

In terms of graphics processing, the device includes a Blackwell Ultra graphics processing unit equipped with 252 gigabytes of dedicated HBM3e video memory, offering a substantial data transfer bandwidth of 7.1 terabytes per second. This system allows the graphics unit to access a shared memory pool of up to 748 gigabytes, delivering computing power of up to 20 petaflops in FP4 processing tasks.

HP provides additional flexibility for users by enabling the integration of separate Nvidia RTX Pro graphics cards through a high-speed PCIe Gen 5 x16 slot, along with two additional PCIe Gen 5 x8 slots and four dedicated M2 slots for high-speed storage. To connect multiple devices in advanced computing clusters, the workstation supports the Nvidia ConnectX 8 network card, offering exceptional transfer speeds of up to 800 gigabits per second through two QSFP112 ports.

The interface features a wide array of ports, including six USB 3.2 Gen 2 Type A ports, two USB 3.2 Gen 2 Type C ports, a 3.5 mm audio port, dedicated audio and microphone ports, as well as two wired Ethernet ports with speeds of 1 gigabit and 10 gigabits per second. Although HP has not disclosed an official price for the workstation yet, estimates suggest it may reach around 100000 dollars based on similar competitive hardware.