Intel Gaudi 2 Dual-OAM Baseboard Reference Platform

The Intel Gaudi 2 Dual-OAM Baseboard Reference Platform (HLRD-225-2) is a high-density AI accelerator baseboard designed to house two Intel Gaudi 2 OAM (Open Accelerator Module) compute modules within a single chassis-ready form factor. Built on Intel's second-generation Gaudi architecture, the platform delivers substantial deep learning training and inference throughput through its integrated high-bandwidth memory, dedicated matrix multiplication engines, and a fully non-blocking on-chip network fabric. The baseboard exposes dual 100GbE RDMA-capable Ethernet ports per Gaudi 2 module, enabling low-latency scale-out clustering without proprietary interconnect dependencies.

Each Gaudi 2 OAM module integrated into this reference platform features 96 GB of HBM2e memory across the two accelerators combined, along with 24 dedicated Tensor Processor Cores (TPCs) and a Matrix Multiplication Engine (MME) per die, optimized for mixed-precision workloads including BF16 and FP32 operations. The platform conforms to the OAM specification under the Open Compute Project (OCP), promoting hardware interoperability and vendor flexibility in large-scale AI infrastructure deployments. Thermal management is handled via direct liquid cooling compatibility, supporting modern high-density datacenter rack architectures.

The HLRD-225-2 is positioned as a reference design for system integrators and datacenter operators building scalable AI training clusters for large language models, computer vision pipelines, and recommendation systems. Its dual-module architecture allows enterprises to evaluate and validate Gaudi 2 performance at reduced rack footprint before scaling to full eight-accelerator OAM chassis configurations. The platform is supported by Intel's SynapseAI software stack, which provides framework integration with PyTorch and TensorFlow, enabling straightforward workload portability from existing GPU-based infrastructure.

Ideal for

• Large language model (LLM) pre-training and fine-tuning in enterprise AI research environments requiring scalable, high-throughput accelerator clusters
• Computer vision model training for manufacturing quality inspection, medical imaging analysis, and autonomous systems development
• Datacenter AI inference serving for real-time natural language processing, recommendation engines, and fraud detection pipelines
• Reference architecture validation and proof-of-concept deployments for organizations evaluating Gaudi 2 before full-scale OAM chassis procurement
• High-performance deep learning research workloads in academic institutions and national laboratories requiring open, standards-based accelerator hardware
• Scalable generative AI infrastructure build-out across UAE, GCC, EMEA, and APAC datacenter environments requiring globally available, enterprise-grade AI compute

Technical specifications

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