HPE AMD Instinct MI300X 192GB OAM GPU Accelerator

The HPE AMD Instinct MI300X 192GB OAM GPU Accelerator (P64568-B21) is built on AMD's CDNA 3 architecture, combining CPU and GPU compute dies with high-bandwidth memory in a unified chiplet design. The accelerator integrates 192GB of HBM3 memory across eight stacks, delivering up to 5.3 TB/s of aggregate memory bandwidth — a defining advantage for memory-bound AI inference and large-scale model training workloads. The OAM (OCP Accelerator Module) form factor ensures compatibility with open-standard AI accelerator platforms and high-density server configurations designed for modern datacenter deployments.

With 304 compute units and a peak FP16/BF16 throughput of 1.307 PFLOPS, the MI300X is engineered to handle the most demanding generative AI, large language model (LLM), and high-performance computing (HPC) tasks at enterprise scale. The massive unified memory pool eliminates the need for model sharding across multiple accelerators for many frontier-scale models, reducing system complexity and improving inference latency. Support for FP8, FP16, BF16, FP32, FP64, and INT8 precisions provides flexibility across training and inference pipelines.

As an HPE-branded component, the MI300X OAM accelerator is validated and integrated into HPE's enterprise server and AI infrastructure portfolio, including HPE Cray XD and ProLiant platforms designed for AI factories and exascale-class compute environments. The ROCm open software ecosystem provides a comprehensive, vendor-neutral foundation for AI/ML frameworks including PyTorch, TensorFlow, and JAX, enabling enterprises to build and scale AI workloads without proprietary software lock-in.

Ideal for

• Large language model (LLM) training and fine-tuning, where the 192GB unified HBM3 memory pool accommodates multi-billion parameter models on fewer accelerators
• Generative AI inference serving for production deployments requiring low latency and high throughput across text, image, and multimodal models
• High-performance computing (HPC) simulations in scientific research, computational fluid dynamics, and molecular dynamics where FP64 double-precision performance is critical
• Enterprise AI factory buildouts requiring open-standard OAM form factor accelerators within high-density, scalable rack architectures
• Multi-node distributed training workloads leveraging high-bandwidth interconnects across HPE Cray XD supercomputing platforms
• Data analytics and numerical computing pipelines requiring sustained memory bandwidth and broad floating-point precision support across mixed-precision workloads

Technical specifications

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