HPE Expands Liquid-Cooled Supercomputing Portfolio, Adds AI-Optimised Servers for Large Enterprises

Hewlett Packard Enterprise (HPE) is raising the bar for supercomputing and artificial intelligence (AI) infrastructure with its latest innovations. Announced on 14 November 2024, HPE unveiled a new portfolio of high-performance computing (HPC) and AI systems, designed to accelerate scientific discovery and empower organisations tackling complex AI workloads, from training large language models to sovereign AI initiatives.

The new offerings feature groundbreaking liquid-cooling technologies and advanced computing power from AMD, Intel, and NVIDIA, underscoring HPE’s commitment to performance, efficiency, and customisation.

A Revolution in Cooling: 100% Fanless Liquid Cooling

At the heart of HPE’s expanded portfolio is its 100% fanless direct liquid cooling system, an industry first. Designed for its Cray EX Supercomputing line, this cooling architecture promises energy efficiency while reducing costs for organisations running intensive workloads.

HPE’s Cray Supercomputing solutions now span compute nodes, networking, and storage, bolstered by updated software to optimise system efficiency and power management.

Next-Generation Supercomputing Innovations

Key innovations in HPE’s Cray portfolio include:

• Cray EX4252 Gen 2 Compute Blade: Featuring eight AMD EPYC processors, it delivers up to 98,304 cores in a single cabinet. Available in Spring 2025, it offers unmatched density and performance for demanding computational tasks.
• Cray EX154n Accelerator Blade: With up to 224 NVIDIA Blackwell GPUs per cabinet, this blade is built for extreme workloads like multi-modal AI training. It’s set to launch by the end of 2025.
• HPE Slingshot Interconnect 400: Doubling line speeds to 400Gbps, this interconnect ensures faster, more efficient communication across large-scale clusters.
• Cray Storage Systems E2000: Delivering double the I/O performance of its predecessor, it supports faster data access and utilisation across CPU and GPU compute nodes.
• Cray User Services Software: Already available, this software improves the user experience, optimising workload management and energy efficiency.

AI-Optimised HPE ProLiant Compute XD Servers

In addition to its supercomputing offerings, HPE introduced two new servers optimised for AI model training and deployment:

• HPE ProLiant Compute XD680 Server: Designed for cost-effective AI workloads, it features Intel Gaudi 3 accelerators and will be available in December 2024.
• HPE ProLiant Compute XD685 Server: Built for performance-driven AI training, this server supports NVIDIA GPUs and HPE’s liquid-cooling capabilities. Versions featuring NVIDIA H200 GPUs will debut in early 2025, while AMD-powered variants are already available.

These servers are aimed at service providers and enterprises scaling their AI capabilities. With HPE’s support for customisation, validation, and testing, customers can expedite deployment and ensure seamless integration into existing systems.

Pushing Boundaries in AI and Supercomputing

The demand for HPC and AI infrastructure has surged as enterprises and governments accelerate sovereign AI initiatives and large-scale model training. Trish Damkroger, HPE’s Senior Vice President and General Manager of HPC & AI Infrastructure Solutions, highlighted the company’s role in meeting this demand:

“Service providers and nations investing in sovereign AI initiatives are increasingly turning to high-performance computing as the critical backbone enabling large-scale AI training that accelerates discovery and innovation.”

Looking Ahead

HPE’s expanded portfolio will be showcased at SC24 in Atlanta from 17–22 November, offering attendees a closer look at its next-generation supercomputing solutions.

As the boundaries of AI and HPC continue to expand, HPE’s innovations position it as a leader in delivering tailored, high-performance systems for organisations pushing the limits of technology.