HPE has released more information about the technical specifications of its forthcoming Cray Supercomputing GX5000 platform.

The company also announced that the High-Performance Computing Center of the University of Stuttgart (HLRS) and Leibniz Supercomputing Center (LRZ), both in Germany, have selected the GX5000 platform for their next-generation supercomputers – Herder and Blue Lion, respectively.

First announced in late October 2025, HPE said its GX5000 platform utilizes AI and high-performance computing (HPC) architecture to “streamline operations site-wide and across distributed clusters.” GX5000-based systems are expected to start being delivered from early 2027.

This week, the company provided further technical details about the platform’s compute, storage, and networking specifications, including the three new processing blades that can be deployed in GX5000 systems.

The HPE Cray Supercomputing GX440n Accelerated Blade supports mixed-precision computing with four Nvidia Vera CPUs and eight Nvidia Rubin GPUs. Up to 24 of these blades can be configured per compute rack, scaling up to 192 Rubin GPUs per rack.

The HPE Cray Supercomputing GX350a Accelerated Blade, meanwhile, supports AMD hardware deployments and includes one AMD Venice Epyc CPU and four AMD Instinct MI430X GPUs. Up to 28 of these blades can be configured per compute rack, providing up to 112 MI430X GPUs per rack.

Finally, the HPE Cray Supercomputing GX250 Compute Blade is comprised of eight AMD Venice Epyc CPUs and is designed for customers looking to build a CPU-only partition. HPE said this blade can be built on either of the other processing blades, with up to 40 able to be configured per compute rack to deliver x86 CPU core density.

All three blades are 100 percent liquid-cooled, and HPE said a mix of the three processing blades can be deployed in a GX5000 compute rack, enabling customers to configure the optimal mix of blades for their workload.

The system’s K3000 storage system, based on the HPE ProLiant Compute DL360 Gen12 server, delivers “industry-leading compute performance, memory density with scalability and high-speed data transfer rates to run demanding applications.” Depending on system requirements, the storage servers will be available in multiple drive sizes and configurations, including eight, 12, or 16 NVMe drives and 512 GB, 1,024 GB, or 2,048 GB DRAM configurations.

HPE has also made its Slingshot 400 available for GX5000-based systems, and the company said the new chassis that has been developed for the blades features 64 400Gbps ports, which can be deployed in eight, 16, or 32 switch configurations with 512, 1,024, and 2,048 ports, respectively.

HPE Slingshot 200, InfiniBand NDR, and 400Gbps Ethernet are also available as connectivity options.

“The latest HPE Cray supercomputing platform reflects the deep collaboration between AMD and HPE to develop leadership technologies at the convergence of HPC and AI,” said Travis Karr, corporate vice president, HPC and sovereign AI at AMD. “By bringing together high-density compute blades powered by AMD Epyc CPUs and AMD Instinct GPUs, we’re helping customers tackle some of the world’s most complex scientific and AI challenges with scalable, energy-efficient solutions.”