HPE ProLiant XL675d Gen10 Plus AMD MI250X 8-GPU Node

The HPE ProLiant XL675d Gen10 Plus AMD MI250X 8-GPU Node (P42138-B21) is a purpose-built high-density compute node designed for the most demanding AI training, HPC simulation, and large-scale machine learning workloads in modern data centers. Integrating eight AMD Instinct MI250X accelerators — each built on the CDNA2 architecture and delivering up to 383 TFLOPS of FP16 peak compute — the node provides exceptional throughput for tensor operations, deep learning model training, and scientific computing. AMD's second-generation CDNA architecture underpins the MI250X, offering hardware support for matrix operations via the Matrix Core Engine, alongside high-bandwidth memory and a unified memory programming model through ROCm.

Each MI250X OAM module in this node contains two Graphics Compute Dies (GCDs) interconnected via AMD Infinity Fabric, effectively presenting as two logical GPUs per physical package. The node's eight OAM slots accommodate up to 16 logical GPU instances, with each GCD carrying 64 GB of HBM2e memory for a combined GPU memory capacity of 128 GB per MI250X module and up to 1 TB of aggregate HBM2e across the fully populated node. Inter-GPU communication is handled through high-bandwidth AMD Infinity Fabric links, enabling low-latency collective operations critical for distributed training and tightly coupled parallel workloads.

Hosted in the HPE ProLiant XL675d Gen10 Plus chassis within the Apollo 6500 Gen10 Plus platform, this node is engineered for dense rack deployment and integrates with HPE's broader management ecosystem, including iLO and HPE OneView, for streamlined provisioning and monitoring at scale. The platform supports AMD EPYC 7003 series processors and PCIe Gen4 connectivity, ensuring balanced CPU-to-GPU bandwidth for data-intensive pipelines. This solution targets enterprise data centers, national laboratories, cloud service operators, and research institutions across regions including the UAE, GCC, EMEA, and APAC seeking to accelerate AI and HPC infrastructure at scale.

Ideal for

• Large-scale deep learning model training for natural language processing, computer vision, and generative AI across distributed multi-node GPU clusters
• High-performance computing simulations in computational fluid dynamics, molecular dynamics, and climate modeling requiring massive parallelism
• AI inferencing at scale for real-time and batch inference serving of large foundation models in enterprise production environments
• Financial services workloads including Monte Carlo risk simulations, portfolio optimization, and quantitative analytics requiring high FP64 throughput
• Seismic processing and reservoir simulation for energy sector organizations demanding sustained double-precision floating-point performance
• Government and academic research computing for physics simulations, genomics pipelines, and graph analytics on sovereign or on-premises HPC infrastructure

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.