Hewlett Packard Enterprise Expedites AI Training with Turnkey Solution

Hewlett Packard Enterprise (HPE) has introduced a supercomputing solution tailored for generative AI applications, specifically aimed at large enterprises, research institutions, and government organizations. This innovative system is designed to expedite the training and fine-tuning of artificial intelligence (AI) models, leveraging private datasets for enhanced performance. The comprehensive turnkey solution encompasses a sophisticated software suite, empowering users to efficiently train, fine-tune, and develop AI applications.

Notably, it integrates liquid-cooled supercomputers, accelerated computers, advanced networking components, and robust storage capabilities. Furthermore, HPE provides accompanying services, forming a holistic package that accelerates organizations' ability to unlock and derive value from AI applications. This strategic initiative signifies HPE's commitment to advancing AI capabilities for diverse sectors, fostering quicker realization of the transformative potential inherent in artificial intelligence.

“The world’s leading companies and research centers are training and tuning AI models to drive innovation and unlock breakthroughs in research, but to do so effectively and efficiently, they need purpose-built solutions,” said Justin Hotard, executive vice president and general manager, HPC, AI & Labs at Hewlett Packard Enterprise. “To support generative AI, organizations need to leverage solutions that are sustainable and deliver the dedicated performance and scale of a supercomputer to support AI model training. We are thrilled to expand our collaboration with NVIDIA to offer a turnkey AI-native solution that will help our customers significantly accelerate AI model training and outcomes.”

HPE's Supercomputing Solution for Generative AI with Unprecedented Scalability and Performance

The supercomputing turnkey solution for generative AI introduced by Hewlett Packard Enterprise (HPE) encompasses pivotal software tools crucial for constructing AI applications, tailoring pre-built models, and coding development and modification. A significant integration feature is the alignment with HPE Cray supercomputing technology, leveraging the robust architecture synonymous with the world's fastest supercomputer, and empowered by NVIDIA Grace Hopper GH200 Superchips. This amalgamation presents organizations with unparalleled scalability and performance essential for handling substantial AI workloads, including the training of large language models (LLM) and deep learning recommendation models (DLRM).

The synergy between HPE's Machine Learning Development Environment and this supercomputing solution is exemplified by the swift fine-tuning of the open-source 70 billion-parameter Llama 2 model in under 3 minutes. This achievement directly translates to accelerated time-to-value for customers. Notably, the advanced supercomputing capabilities of HPE, bolstered by NVIDIA technology, result in a remarkable 2-3X enhancement in system performance, further underlining the solution's prowess in advancing AI capabilities. This strategic integration signifies a significant leap forward in the efficiency and effectiveness of AI development and implementation for organizations leveraging this cutting-edge solution.

“Generative AI is transforming every industrial and scientific endeavor,” said Ian Buck, vice president of Hyperscale and HPC at NVIDIA. “NVIDIA’s collaboration with HPE on this turnkey AI training and simulation solution, powered by NVIDIA GH200 Grace Hopper Superchips, will provide customers with the performance needed to achieve breakthroughs in their generative AI initiatives.”

A powerful, integrated AI solution

The generative AI supercomputing solution is a meticulously designed, unified, AI-centric package encompassing the following comprehensive technologies and services from end to end:

–  AI/ML Acceleration Software: A set of three software tools designed to assist users in training and fine-tuning AI models and developing customized AI applications.

• HPE Machine Learning Development Environment is a software platform for machine learning (ML) that expedites the development and deployment of AI models. By seamlessly integrating with popular ML frameworks and simplifying data preparation, it enables customers to create and implement AI models more efficiently.
• NVIDIA AI Enterprise propels organizations into the forefront of AI, emphasizing security, stability, manageability, and support. This solution provides an array of extensive frameworks, pre-trained models, and tools that streamline the development and deployment of AI in production environments.
• The HPE Cray Programming Environment suite equips programmers with a comprehensive set of tools for developing, porting, debugging, and refining code. This suite offers a holistic solution for optimizing the programming process across various stages, enhancing the efficiency and effectiveness of code development.

– Designed for Scalability – Built upon the HPE Cray EX2500, an exascale-class system, and equipped with cutting-edge NVIDIA GH200 Grace Hopper Superchips, this solution boasts the capability to scale up to thousands of graphics processing units (GPUs). It excels in dedicating the complete capacity of nodes to accommodate a single, high-impact AI workload, ensuring a swift time-to-value. Distinguished as the inaugural system to showcase the quad GH200 Superchip node configuration, it represents a pioneering advancement in scalable computing for AI applications.

– Real-time AI Networking - The HPE Slingshot Interconnect introduces a high-performance network based on open, Ethernet principles, tailored to support exascale-class workloads. Grounded in HPE Cray technology, this adaptable interconnection significantly enhances system-wide performance by facilitating exceptionally high-speed networking.

– Simplified Turnkey Experience - Enhanced by HPE Complete Care Services, the solution includes global specialists for seamless set-up, installation, and comprehensive lifecycle support, streamlining the adoption of AI.

The future of supercomputing and AI will be more sustainable

By 2028, it is estimated that the growth of AI workloads will require about 20 gigawatts of power within data centers. Customers will require solutions that deliver a new level of energy efficiency to minimize the impact of their carbon footprint.

By 2028, the expansion of AI workloads is projected to necessitate approximately 20 gigawatts of power within data centers. In response, customers will seek solutions that provide heightened energy efficiency to mitigate the environmental impact of their carbon footprint.

At the heart of HPE's computing initiatives lies a commitment to energy efficiency. HPE's solutions incorporate liquid-cooling capabilities, achieving up to a 20% performance improvement per kilowatt compared to air-cooled alternatives, while consuming 15% less power.

HPE holds a distinctive position to assist organizations in harnessing cutting-edge computer technology to propel their AI objectives while concurrently aiding in the reduction of energy consumption.

Availability

The generative AI supercomputing solution will be widely accessible through HPE in over 30 countries starting in December.