Introduction

Hewlett Packard Enterprise has a new server family built for AI workloads.

What HPE is actually shipping

The headline spec is density. The Gen13 lineup supports up to 256 cores per socket with high-core-count EPYC chips, and up to 600W of TDP per socket.

AI and Memory Upgrades

Memory gets a serious upgrade too. The servers support up to 16 memory channels of DDR5 memory running at up to 8,000 MT/s.

Four machines, four jobs

The ProLiant DL525 Gen13 is designed for CPU-based AI inference. Inference is the part of AI where a trained model actually answers questions or makes predictions, as opposed as the expensive training phase.

Accelerator-Optimized and Advanced Cooling

The ProLiant DL585a Gen13 is accelerator-optimized and supports up to eight double-wide GPUs.

Hybrid Workloads

Rounding out the lineup are the XD245 and XD285, which focus on advanced cooling for CPU-intensive HPC and AI workloads.

Security Features

HPE says the lineup is built for hybrid workloads, spanning agentic AI, inference, analytics and modeling. Agentic AI refers to systems that can take actions and complete multistep tasks on their own, rather than simply responding to prompts.

Quantum-Ready Security

The new version of HPE iLO 8 adds automatic self-encrypting drive support, so data on storage is encrypted without someone having to remember to switch it on. It also includes readiness for post-quantum cryptography.

Background

HPE says it applied lessons from its work on exascale supercomputers Frontier and El Capitan. Exascale systems can perform at least a quintillion calculations per second.

Launch and Partnership

The launch also follows AMD's earlier announcements around its EPYC 9006 series. HPE's rollout puts that silicon into production systems aimed squarely at enterprise buyers.