Accelerate your AI workloads with flexible access to GPU clusters at scale

[HPE_Experts]

Running into performance walls with cloud-based AI workloads? Discover how HPE GPU Cloud Service offers scalable, high-performance compute that simplifies AI development — without the infrastructure hassle.

It’s no secret that the growth we are seeing in AI today is unprecedented. Model training, fine-tuning, and inference have become commonplace across various industries, empowering organizations of all sizes to use AI for productivity gains and valuable insights. Our customers frequently share their excitement about the incredible opportunities for innovation and swift market entry. 

However, alongside these benefits come new challenges. To meet the demands of rapid data processing, intensive AI model training, and complex high-performance computing (HPC) workloads, organizations require powerful, scalable, and flexible computing resources. The decision on how best to utilize this power — whether on-premises, in the cloud, or through a hybrid model — depends on several factors. 

A recurring challenge we hear from our customers is the limited options for consuming HPC in the cloud. Many have noted performance limitations compared to on-premises systems — or even common cloud service providers — especially with data-intensive AI workloads that demand strong scaling and reliable performance. Our customers need the flexibility to scale as demands increase while maintaining top performance and cost efficiency. 

With the expansion of the HPE portfolio, we now offer an alternative to running AI workloads in the public cloud: HPE GPU Cloud Service. This service is designed for GPU-accelerated workloads, enabling organizations to access top-performing computing power to accelerate AI and machine learning applications. The flexibility to access accelerator hardware and support multiple workloads allows enterprises to achieve rapid time-to-insights and unlock new business value through deeper, faster, and more accurate predictions — without the need for significant upfront investment.

Realizing the benefits of GPU acceleration 

AI and ML workloads demand immense processing power, particularly for training large models or handling large-scale, latency-sensitive inferencing. Traditional CPUs often fall short of these demands, making GPUs the preferred choice for many AI applications. GPUs significantly accelerate computations, enabling businesses to achieve faster results and make groundbreaking advancements. However, setting up GPU infrastructure in-house can be costly and resource-intensive, especially as workloads fluctuate. HPE GPU Cloud Service eliminates these barriers, allowing organizations enterprises to flexibly harness high-performance GPUs. This service simplifies the process and helps organizations unlock new business value. 

Beyond just infrastructure, HPE GPU Cloud Service offers more than bare metal capabilities. The cloud provides a fully managed Kubernetes experience with an RKE2-based Kubernetes API and an out-of-the-box Rancher GUI. Additionally, a suite of performance-optimized cloud services and commonly used open-source applications, drivers, and tools are available for rapid deployment. This allows customers to focus on managing workloads and services rather than configuring the environment. 

HPE GPU Cloud Service: accelerated access with a trusted partner 

Powered by HPE’s portfolio of industry-leading platforms, HPE GPU Cloud Service is architected with HPE’s supercomputing expertise, providing an ideal platform for GPU-accelerated workloads. This service enables large enterprises to develop and train AI models faster, speeding time-to-results with unprecedented efficiency and precision, leading to detailed insights. Designed to simplify access to HPC, it offers a fully managed cloud with built-in automation to run customer workloads. 

HPE GPU Cloud Service is currently based on the HPE Cray XD670, a system that has achieved the #1 spot for Natural Language Processing in recent MLPerf inference benchmark results from MLCommons.

Source: “New MLPerf Inference Benchmark Results Highlight The Rapid Growth of Generative AI Models,” MLCommons, March 2024

Our cloud-based model provides key advantages for large enterprises looking to accelerate AI workloads. Customers without existing scalable, high-performance GPU capabilities can now access compute resources in the HPE cloud rather than investing in and managing on-premises infrastructure.  

 Alternatively, customers who already own and operate high-performance systems for core workloads can meet the changing demands of new workloads with a cloud service model from HPE. Adding HPE GPU Cloud Service can help improve ROI from computing while maintaining cost predictability. 

 Another benefit is the flexibility HPE GPU Cloud Service provides to organizations with varying needs. For example, a customer may need a system with 256 GPUs for three years and make the resource available to various teams and projects. HPE GPU Cloud Service can provide access to high-performance systems without relying on internal HPC system management expertise or constrained resources. Our tightly coupled HPC systems deliver the fastest possible performance at scale, with the flexibility to create, manage, and allocate clusters to users quickly and easily. 

Why choose HPE for GPU-accelerated workloads? 

HPE’s commitment to AI-driven innovation enables customers to access GPUs, backed by a trusted, industry-leading partner. With a proven track record in HPC, HPE provides the expertise and infrastructure to support even the most demanding workloads. 

Partnering with HPE for GPU-accelerated workloads provides enterprises with: 

• Flexible access to HPC: Direct access to accelerator hardware and support for multiple workloads and a diverse set of users. 

• Scalable performance: Top performance with the latest CPUs, GPUs, interconnects, storage, and software. 

• Kubernetes clusters: Bring containerized workloads to an AI and HPC supercomputer. 

• Cost efficiency: Maximize investments with a cloud-based consumption model; no data ingress or egress fees on up to 100Gbps internet connection. 

• Trusted partner: Innovate with HPC solutions engineered to work together efficiently from the industry leader in supercomputing, with a commitment to sustainability and expert support throughout the entire lifecycle. 

HPE GPU Cloud Service is a game-changer for AI and HPC workloads. By providing flexible, cloud-based access to powerful GPU resources, HPE enables large enterprises to accelerate AI initiatives, drive innovation, and maintain cost control. With HPE as a trusted partner, organizations can tackle the challenges of AI and HPC with confidence, unlocking new potential for growth and transformation. 

For more information on HPE GPU Cloud Service, view the data sheet or contact the HPE GPU Cloud Service organization at GPUs@hpe.com. 

---

Meet Maryam Chaudry, VP & GM of AI Cloud, HPE

Maryam Chaudry is the Vice President & General Manager of AI Cloud at HPE. She has been with HPE for 19 years – starting her career at the company as an intern – and has held a variety of roles in product management, category management, business strategy and planning, pricing, and go-to-market. Maryam is an alumna of the University of Houston and a passionate advocate of women’s education with a keen interest in supporting women in STEM. She is actively involved through mentorship, community involvement, and supporting organizations that nurture and cultivate the next generation of intelligent and outstanding leaders.