Detailed Specifications
HPE P18065-B21 32GB NVIDIA PCIe Accelerator
Overview
The HPE P18065-B21 is a 32 GB computational accelerator based on an NVIDIA architecture, designed to offload intensive parallel workloads such as AI/ML training, HPC, and data analytics. It connects via PCI Express to HPE ProLiant and Apollo systems, providing high-throughput, low-latency GPU compute. This card is used in data center compute nodes, particularly those running AI or simulation workloads, to accelerate tasks efficiently. It integrates with HPE server ecosystems and supports standard data center management tools.
Key Features
32 GB high-bandwidth GPU memory for compute-intensive tasks
PCI Express x16 interface for maximum data transfer
High parallelism via CUDA cores for AI, HPC, and rendering
Supported by NVIDIA CUDA, cuDNN, TensorRT frameworks
Passive cooling design (requires server airflow)
Technical Specifications
| Specification | Details |
|---|
| Brand | HPE |
| Part Number | P18065-B21 |
| Product Type | Computational GPU Accelerator |
| Processor Installed | NVIDIA GPU (model not publicly disclosed) |
| Processor Supported (Max) | Same on-chip GPU |
| Memory Installed | 32 GB (HBM or GDDR) |
| Memory Slots | Not applicable |
| Maximum Memory Supported | 32 GB |
| Drive Bays | Not applicable |
| Storage Controller | Not applicable |
| Network Adapter | Not applicable |
| Form Factor (H × W × D in inches) | Approx. 4.37 × 10.5 in (full-height, dual-slot PCIe) |
| Weight (lbs) | Approx. 2.20 lbs |
| Power Supply / Consumption | ~250 W (TDP) |
| Rack Units | Not applicable |
| Cooling Method | Passive heatsink, needs system airflow |
| Supported APIs / Frameworks | CUDA®, cuDNN, TensorRT |
