Skip to main content
New · Expected March 2027

HPE ProLiant DL585a Gen13 Server

Production AI with exceptional GPU density. Dual AMD EPYC™ SP7 processors and up to eight double-wide GPUs in a 10U platform maximize GPU utilization and token throughput, with flexible GPU, cooling, power, and fabric options that fit existing data centers and scale with demand.

Request a Quote Download the Datasheet

HPE lists the DL585a Gen13 for availability in March 2027. Request a quote now and a specialist will help you plan GPU, power, and cooling ahead of launch.

HPE ProLiant DL585a Gen13 10U GPU server
8
Double-wide PCIe GPUs
2 × 256
Cores across dual AMD EPYC™ 9006 CPUs
Up to 8 TB
DDR5 across 32 DIMM slots
N+N
Dual power domains, up to 14 × 3200W

Engineered to scale AI with exceptional GPU density

Are you looking for a server engineered to scale AI workloads with exceptional GPU density and performance?

The new HPE ProLiant DL585a Gen13 Server, supporting dual AMD EPYC™ SP7 processors and up to eight double-wide GPUs in a 10U format, is purpose-built for production AI. By maximizing GPU utilization and token throughput, it helps organizations operationalize AI inference and visualization. Designed with flexible GPU, cooling, power, and fabric options, the DL585a meets evolving workload and data center requirements today and in the future, with advanced security, comprehensive infrastructure management tools, and intelligent automation to deploy and operate AI infrastructure with confidence.

What's New

DL585a Gen13 at a glance

Form factor10U rack
ProcessorsTwo AMD EPYC™ 9006 SP7 Server CPUs, 64 to 256 cores each, up to 5 GHz maximum and up to 1024 MB L3 cache, depending on processor
Memory32 DDR5 DIMM slots (16 per processor, 16 channels per processor): up to 8 TB with 256 GB RDIMMs at 8000 MT/s, or up to 4 TB with 128 GB MRDIMMs at 12800 MT/s
Memory protectionAdvanced ECC, online spare, mirroring, combined channel (lockstep), and HPE Fast Fault Tolerant Memory (ADDDC)
AcceleratorsUp to eight qualified double-wide PCIe GPUs (0, 1, 2, 4, or 8 GPU configurations), including NVIDIA RTX PRO 6000 Blackwell Server Edition; NVIDIA 4-way NVLink Bridge support
ExpansionUp to 8 PCIe Gen6 x16 expansion slots and 1 OCP 3.0 slot, depending on configuration
StorageUp to 16 EDSFF E3.S NVMe drives; optional HPE NS204mff OS boot device
NetworkingNo embedded networking; an OCP 3.0 or PCIe stand-up networking adapter is required
PowerUp to 14 HPE 3200W M-CRPS power supplies (up to 96% efficiency): 1+1 redundancy for the system board and N+N redundancy for GPUs
Cooling15 hot-plug dual-rotor fan assemblies
ManagementHPE iLO 8 Standard with Intelligent Provisioning (embedded), HPE OneView Standard (requires download), HPE Compute Ops Management; optional HPE iLO Advanced and HPE OneView Advanced
SecurityEmbedded HPE TPM 2.0 and intrusion detection kit, plus HPE iLO 8 Auto-SED, multi-person authorization, post-quantum cryptography, and component attestation
Warranty3/3/3: three years of parts, three years of labor, and three years of on-site support coverage
AvailabilityExpected March 2027

Automated protection, verified trust with HPE iLO 8

The DL585a Gen13 with HPE iLO 8 protects data and firmware from first boot and keeps AI infrastructure updated throughout its lifecycle.

  • Auto-SEDStandardizes on self-encrypting drives, automatically protecting data at rest from first boot without additional configuration, external key managers, or manual activation.
  • Multi-person authorizationBuilt-in approval controls for high-impact actions such as key provisioning and secure erase reduce risk from mistakes, unauthorized actions, or credential misuse.
  • Post-quantum cryptographyExtended PQC protection for live communications between HPE iLO, platform firmware, option firmware, embedded components, and management systems.
  • Expanded component attestationAuthenticates an expanded list of critical system components at every stage of the server lifecycle, so nothing operates unless it is trusted.
HPE ProLiant DL585a Gen13 server with HPE iLO 8

Workloads the DL585a was made for

AI inference

Maximize GPU utilization and token throughput to operationalize production inference.

Agentic AI

Dense GPU capacity and high-core-count host CPUs for multi-step, tool-using AI agents.

Retrieval-augmented generation

Pair large memory and fast EDSFF NVMe storage with GPU inference for RAG pipelines.

Fine-tuning

Start with the GPUs you need and scale up to eight double-wide GPUs as models grow.

Analytics

Dual AMD EPYC™ processors and up to 8 TB of DDR5 for data-intensive analytics.

Visualization

GPU-accelerated visualization alongside AI workloads on one ProLiant platform.

Plan your GPU build before the March 2027 launch

GPU count and model, power-domain sizing, networking fabric, storage, and data center power and cooling all need to line up before a DL585a order. Tell us what your AI workloads need to do and we'll help you configure it, price it, and track it through HPE's first ship dates.

  • Configuration guidance from people who know the HPE ProLiant catalog
  • Authorized-partner pricing, without the usual procurement friction
  • Order follow-through, from quote to delivery
Request a Quote

Key features

Dual AMD EPYC™ processors with high core counts deliver the host compute needed to support growing AI and business workloads over time. Flexible AI acceleration lets you start with only the GPUs you need and expand up to eight double-wide PCIe GPUs as workloads increase.

Support for PCIe Gen6 and next-generation fabrics enables smoother adoption of future AI innovations and protects investments as AI infrastructure evolves, while a modular architecture is designed to fit within existing and future data center power and cooling environments.

HPE iLO 8 standardizes on self-encrypting drives through Auto-SED, adds multi-person authorization for high-impact administrative actions, delivers extended post-quantum cryptography (PQC) protection, and expands component attestation across the server lifecycle. Combined with security management, proactive guidance, and direct access to HPE specialists through HPE Tech Care Service, it helps keep servers protected and up to date.

Simplify server onboarding across distributed environments with low-touch operations in HPE Compute Ops Management, and improve day-2 decision making with expert guidance and personalized insights from HPE Tech Care. Closed-loop, AI-enabled workflows and HPE Support integration deliver faster mean time to repair (MTTR), advisory intelligence powered by agentic AI moves remediation from manual reviews to automated workflows, and power capacity data helps plan rack capacity with confidence.

Datasheets & downloads

HPE ProLiant DL585a Gen13 Server Datasheet
Overview, features, and technical specifications (PDF)
HPE ProLiant DL585a Gen13 Server QuickSpecs
Configuration, GPU, power, and ordering detail (PDF)

HPE ProLiant DL585a Gen13: common questions

A 10U server supporting dual AMD EPYC™ SP7 processors and up to eight double-wide GPUs, purpose-built for production AI including inference, agentic AI, RAG, fine-tuning, analytics, and visualization.

Up to eight qualified double-wide PCIe GPUs in 0, 1, 2, 4, or 8 GPU configurations, including the NVIDIA RTX PRO 6000 Blackwell Server Edition. Four neighboring GPUs can be connected with an NVIDIA 4-way NVLink Bridge to boost inference speeds, and the platform supports the NVIDIA 2-8-5 reference architecture.

Two AMD EPYC™ 9006 SP7 processors with 64 to 256 cores each at up to 600W, and 32 DDR5 DIMM slots supporting up to 8 TB with 256 GB RDIMMs at 8000 MT/s or up to 4 TB with MRDIMMs at 12800 MT/s.

HPE lists the DL585a Gen13 for availability in March 2027. ServerComputeWorks is taking quote requests now; contact a specialist to plan and price your configuration.

It uses dual power domains with up to 14 HPE 3200W M-CRPS power supplies: up to two for the system domain and up to 12 dedicated to the GPU domain, fully N+N redundant.

Request a quote for the DL585a Gen13

Expected March 2027. Tell us about your AI inference, agentic AI, RAG, or visualization workloads and a ServerComputeWorks specialist will help you plan, configure, and price the right DL585a Gen13 build.

Request a Quote

*AMD and AMD EPYC are trademarks of Advanced Micro Devices, Inc. NVIDIA and NVLink are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. All other third-party trademark(s) is/are property of their respective owner(s). Specifications and availability are subject to change.