HPE ProLiant DL585a Gen13 Targets Enterprise AI With PCIe Gen6 GPU Density

As of Friday, October 9, 2026, the HPE ProLiant DL585a Gen13 is one of the most interesting systems in HPE’s newly announced ProLiant Gen13 lineup. HPE introduced the first four AMD-powered ProLiant Gen13 servers on October 7, 2026, positioning them for workloads such as enterprise AI inference, agentic AI, analytics, virtualization, simulation, and modeling. The DL585a Gen13 sits at the high-density end of that launch: a 10U PCIe-based GPU platform designed around dual 6th Gen AMD EPYC 9006 server processors, PCIe Gen6 connectivity, and support for up to eight double-wide PCIe GPUs. Availability is being described for March 2027, so this is best viewed as a product overview of a newly announced platform rather than a review of shipping hardware. (hpe.com)

The DL585a Gen13 is built for enterprises that want a dense accelerator server without moving directly into a full rack-scale AI appliance. HPE lists the system with dual AMD EPYC SP7 processors, using the AMD EPYC 9006 server processor generation, with support for CPUs reaching up to 256 cores at 600W. On the accelerator side, the headline figure is support for up to eight double-wide GPUs in the 10U chassis. HPE also notes support for connecting four neighboring GPUs through an NVIDIA 4-way NVLink Bridge, which gives the platform a path for tighter GPU-to-GPU communication in supported configurations. Memory support is listed at 32 DIMM slots with DDR5 speeds up to 8000 MHz and total memory capacity up to 8 TB. For I/O, HPE’s product listing calls out up to eight x16 PCIe Gen6 connections and one OCP slot, giving the server the bandwidth profile expected from a system aimed at inference, RAG model fine-tuning, analytics, and virtualization. (buy.hpe.com)

View NVIDIA DGX Spark Personal AI Desktop Supercomputer on partner website

PCIe Gen6 is not just a spec-sheet upgrade in a server like this. In GPU-heavy systems, the balance between CPU resources, memory bandwidth, accelerator capacity, networking, and storage access can decide how useful the platform is once it lands in a real data center. The DL585a Gen13’s design is aimed at keeping more data moving between CPUs, GPUs, and network adapters, rather than simply adding more accelerators and hoping the rest of the platform keeps up. HPE describes the server as being optimized for GPU utilization, token throughput, and PCIe bandwidth across demanding AI workloads. That framing is important because enterprise AI deployments are increasingly about inference at scale, retrieval-augmented generation, multi-user virtualization, and mixed analytics pipelines, not only about training the largest possible model. (hpe.com)

See ASUS GeForce RTX 5090 price

HPE is also emphasizing the operational side of the DL585a Gen13. The server uses dual power domains, with HPE listing 12 power supplies dedicated to GPUs for N+N redundancy. That matters in accelerator-dense systems because power delivery and serviceability can become just as important as peak compute capability. Management is handled through HPE Integrated Lights-Out 8 and HPE Compute Ops Management, giving IT teams tools for configuration, monitoring, and updates across Gen13 infrastructure. HPE also highlights security features around iLO 8, including post-quantum cryptography protection for live communications between iLO, platform firmware, option firmware, embedded components, and management systems. The company additionally says the platform standardizes on self-encrypting drives through Auto-SED, intended to protect data at rest from first boot without requiring manual activation or an external key manager. (buy.hpe.com)

Buy Crucial T705 13 GB/s NVMe SSD here

The DL585a Gen13 is only one part of HPE’s first ProLiant Gen13 wave. The launch also includes the ProLiant DL525 Gen13 and OCP-style systems such as the ProLiant XD245 and ProLiant XD285, which point toward denser rack-scale deployments and more efficient power and cooling layouts. That wider context is useful: HPE is not presenting Gen13 as a single-purpose AI box, but as a broader compute refresh built around AMD EPYC 9006 CPUs, higher core density, stronger memory bandwidth, modern I/O, automated management, and more deployment options. For organizations evaluating AI infrastructure in late 2026, the DL585a Gen13 stands out because it brings eight-GPU PCIe density into a ProLiant format that enterprise IT teams already understand, while leaving room for different accelerator, networking, and workload choices as deployments mature. (hpe.com)

See Lenovo Legion Pro 7i laptop with Core Ultra 9 and NVIDIA RTX 5090 24GB price