Do you need a highly performant, secure, and energy-efficient large AI cluster? The HPE ProLiant Compute XD685 leverages a purpose-built architecture and is designed to accelerate large, complex AI model training and tuning workloads. Powered by eight of the latest accelerators from NVIDIA or AMD, this server is optimized for large language model training, natural language processing, and multimodal training. Optional direct liquid cooling (DLC) helps organizations meet escalating power requirements, advance sustainability goals, and lower operational costs. Managed by HPE iLO, the server offers seamless management and robust security. HPE Services provides expertise for global installation and quick deployment of large AI clusters, offering stability and operational excellence.

What is new supporting point:
Support for eight NVIDIA Blackwell Ultra (B300 HGX) GPUs, eight NVIDIA Blackwell (B200) GPUs, eight NVIDIA H200 Tensor Core GPUs, or eight AMD Instinct™ MI355X accelerators.
Direct liquid cooling (DLC) available for all GPU options. Air-cooling available for NVIDIA H200 GPUs.
Modular 5U chassis for direct liquid cooling environments or 6U chassis for air cooled configurations.
24x DDR5 6400 RDIMMs and 12 channels per CPU DDR5 with ECC
Seamless management and built-in security features from HPE iLO 6
2x 5th Generation AMD EPYC™ Processors
Alternate Codes : CO12173