The HPE ProLiant Compute XD685 maximizes rack density in a 5U chassis that supports eight nodes per rack for 8-way GPU systems.
Hewlett Packard Enterprise (HPE) has announced the launch of its HPE ProLiant Compute XD685, designed for advanced artificial intelligence (AI) model training, powered by 5th Gen AMD EPYC processors and AMD Instinct MI325X accelerators. This new system is optimized for rapidly deploying high-performance, secure, and energy-efficient AI clusters tailored for large language model (LLM) training, natural language processing, and multi-modal training.
As organizations across sectors such as healthcare, workforce productivity, and climate science strive to unlock AI’s full potential, there is a growing demand for scalable, high-performance solutions that can be deployed quickly. AI service providers, governments, and large-scale model builders require flexible, robust systems to meet these needs.
Trish Damkroger, Senior Vice President and General Manager of HPC & AI Infrastructure Solutions at HPE, highlighted the critical role that scalable computing solutions play in AI advancements. She noted that HPE, a leader in delivering high-performance, energy-efficient systems, has extended its innovation with the HPE ProLiant Compute XD685. This system aims to support the growing AI market and accelerate breakthroughs in scientific and engineering fields.
Forrest Norrod, Executive Vice President and General Manager of AMD’s Data Center Solutions Group, emphasized the strong partnership between HPE and AMD. According to Norrod, the HPE ProLiant Compute XD685, powered by the latest AMD EPYC processors and Instinct accelerators, provides unmatched AI performance and efficiency, helping customers bring AI solutions to market faster and drive industry innovation.
The HPE ProLiant Compute XD685 is engineered to speed up large AI model training and tuning projects. It features a modular 5U chassis that shortens the time to market AI solutions and provides flexibility for a wide range of GPUs, CPUs, components, software, and cooling methods. In addition, HPE’s integration of direct liquid cooling technology and security innovation makes the system both high-performing and secure.
The HPE ProLiant Compute XD685 is engineered to deliver exceptional AI performance while reducing hardware requirements. It is powered by AMD CDNA 3 architecture and AMD Instinct MI325X accelerators, enabling organizations to run the largest AI models with fewer GPUs. These accelerators have HBM3E memory, providing 6 TB/s of memory bandwidth. This ensures faster processing for training and inference tasks, significantly enhancing overall performance while lowering the total cost of ownership (TCO).
For organizations with demanding energy and cooling needs, HPE offers direct liquid cooling to improve efficiency. This cooling solution helps companies meet rising power requirements and sustainability objectives while cutting operational costs. Drawing on HPE’s decades of expertise in deploying some of the world’s largest liquid-cooled IT infrastructures, the ProLiant Compute XD685 utilizes this cutting-edge cooling technology to enhance both performance and environmental efficiency.
To further accelerate AI cluster deployment and streamline operations, HPE provides a comprehensive range of customizable services. These services cover everything from building and integrating solutions to thoroughly testing and customizing them at the factory before deployment. By offering expert guidance throughout this process, HPE ensures that large AI clusters can be deployed anywhere in the world with stability and operational excellence, reducing time to market and enhancing operational outcomes.
Security is also a top priority for HPE, and the ProLiant Compute XD685 offers robust protection through HPE iLO. This integrated management platform ensures that the systems are managed securely from end to end. Security features are built into the hardware, starting at the supply chain level and extending down to the silicon, providing robust protection against growing cybersecurity threats in today’s environment.
The HPE ProLiant Compute XD685 also maximizes rack density in a 5U chassis that supports eight nodes per rack for 8-way GPU systems. This design allows the system to accommodate up to eight AMD Instinct MI325X or MI300X accelerators and two of the latest AMD EPYC 9005 Series processors. Depending on their specific requirements, customers can also choose between air or direct liquid cooling, enabling them to optimize their infrastructure for both performance and efficiency.
HPE Performance Cluster Manager is a fully integrated system management software that simplifies managing large AI clusters. This tool automates the setup of complex systems from bare metal, enabling quick deployment while ensuring clusters remain operational. Detailed telemetry, GPU stress tests, and other advanced features keep AI systems running smoothly, providing customers with a reliable and scalable solution for their most challenging AI workloads.
The HPE ProLiant Compute XD685 can be ordered today through HPE and will be generally available in the first quarter of 2025.
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed
VergeIO launches ioOptimize, leveraging machine learning and narrow AI to extend the life of hardware and improve VergeIO's flexibility. (more…)
Dell PowerMax updated with increased efficiency, improved cyber resiliency, and multicloud mobility to help customers with optimized performance and greater…
Western Digital has announced the availability of two new drives that set industry records in capacity: the 26TB CMR HDD…
Dell Technologies expands GenAI solutions with new offerings from the Dell AI Factory tailored for AMD environments. (more…)
Dell Technologies has unveiled a series of integrated rack-scalable systems, including server, storage, and data management innovations designed for Dell…
Built for data centers, the Eaton heavy-duty enclosures offer flexibility and power to support AI's highest demands. (more…)