Accelerating Deep Learning Insights With GPU-Based Systems
August 3, 2017 Bill Mannel
Explosive data growth and a rising demand for real-time analytics are making high performance computing (HPC) technologies increasingly vital to success. Organizations across all industries are seeking the next generation of IT solutions to facilitate scientific research, enhance national security, ensure economic stability, and empower innovation to face the challenges of today and tomorrow.
HPC solutions are key to quickly answering some of the world’s most daunting questions. From Tesla’s self-driving car to quantum computing, artificial intelligence (AI) is enabling unparalleled compute capabilities and outmatching humans at many cognitive tasks. Deep learning, an advanced AI technique, is growing in popularity as organizations strive to operate faster and smarter than ever. Deep learning applications utilize neural networks to analyze a variety of data types, including machine signals, audio, video, speech, and text. Based on a series of binary true/false questions, this technology can learn to produce conclusions in the same way as humans.
Up to this point, we have only scratched the surface of the deep learning revolution for advanced analytics.
Hewlett Packard Enterprise (HPE) currently offers a comprehensive, purpose-built portfolio of infrastructure hardware, software, and services capabilities to address a broad range of HPC, Big Data, and AI use cases across multiple vertical industries—ranging from financial services, life sciences, manufacturing, energy, government, and academia. Specifically for deep learning, the HPE Apollo 6500 for training, HPE Apollo 2000 for inference engines, HPE Edgeline for edge analytics and inference, and HPE Pointnext advisory and support services enable pervasive analytics from the core data center to the intelligent edge.
HPE has announced a comprehensive GPU-based solutions portfolio tailored to a new frontier of deep learning capabilities. Leveraging robust technologies from our acquisition of SGI, HPE’s enhancements to its existing capabilities include the HPE SGI 8600 (formerly the SGI ICE XA), as well as purpose-built HPC systems in the HPE Apollo portfolio designed for maximum performance, scale, and efficiency. New HPE deep learning innovations create a dense GPU environment, powered by integrated solutions and service offerings to promote automated, real-time intelligence and insights.
At HPE, our objective is to drive performance in all sectors, equipping organizations with the right technologies to harness actionable intelligence, optimize flexible capacity for analytics, and extend compute resources from the core of the data center to the intelligent edge. And we aren’t alone.
Partnering For Success
Building on a strong collaboration with Nvidia, we are proud to introduce new HPE compute platforms and solutions that are optimized for a range of data-rich applications and complex IT workloads. Together, HPE and Nvidia are driving major improvements to system design, IT solutions, and support services that can not only improve compute performance and reduce time-to-value, but also increase financial flexibility for deep learning adoption.
Spearheading these system advancements, the HPE SGI 8600 platform designed for large, petaFLOPS-scale deep learning environments provides optimal cooling and power efficiency to scale up to thousands of nodes. Combined with Nvidia Tesla GPU accelerators featuring high-speed NVLink interconnects and next-generation Nvidia Volta GPUs, the 8600 utilizes GPU-to-GPU communication which enables 10X the FLOPS per node compared to processor only based nodes. HPE and Nvidia are making Volta-based Nvidia Tesla SXM2 GPU-powered Apollo 6500 systems available through an early access program. Customers who want to take early advantage of a platform geared for up to 8 SXM2 Volta GPUs in an NVLinked architecture, which will be one of the sweet spots of deep learning training, will be able to acquire this technology in 4Q 2017.
HPE and Nvidia are also working with Kinetica to enable secure credit card transactions with real-time fraud detection capabilities. This solution optimizes Kinetica’s GPU-based analytics software for in-database processing performance and efficiency for large datasets, which safeguards sensitive customer information and dramatically reduces the risk of data loss without requiring costly upgrades.
Finally, our HPE Pointnext services innovation will ensure world-class global support and lifecycle management for AI and deep learning environments—including HPE SGI 8600 platforms. Pointnext promises to increase financial flexibility, streamline IT transformations, and help customers achieve better business outcomes. Among these services, HPE Flexible Capacity enables organizations to pay as they go while rapidly improving and accelerating business operations.
Long-Term Benefits Of Deep Learning
With the most comprehensive portfolio across all compute, storage, networking, software, and services, HPE platforms are opening the door to the next generation of IT innovation. Our approach to GPU and deep learning enables customers to derive real-time insights from vast quantities of data. And with the largest deployment of AI technologies, we are equipped to meet unique deep learning requirements with superior performance and scalability.
GPU-based solutions are empowering more informed, real-time decisions: “With the need to embed more intelligence and automation into data analytics to address scientific and business challenges, artificial intelligence-based techniques are growing in importance,” says Steve Conway, senior vice president of research at Hyperion Research. “HPE’s systems and solutions innovations announced today address key performance and expertise constraints affecting deep learning. HPE’s Nvidia partnership for deep learning and comprehensive infrastructure capabilities, from the core data center to the intelligent edge, are designed to use automated intelligence to enable real-time insights for customers.”
In conjunction with these solutions, HPE is promoting enhanced Centers of Excellence (CoEs) to assist and guide customers with code modernization, deployment, and infrastructure options. Global HPE/Nvidia CoE capabilities will target GPU acceleration of deep learning, HPC, and advanced visualization workloads, driving innovations that leverage next-generation Nvidia products and support proof of concepts (POCs) and benchmarking. To further empower customers, the Deep Learning Institute will also deliver training and education on deep learning capabilities.
At HPE, we have a strong commitment to GPU computing development and our partner ecosystem. Our goal is to enable our customers for their continued growth and success with industry-leading technologies together with the training, assets, and expertise of our partner community.
To learn more about our comprehensive deep learning portfolio, I invite you to follow us on Twitter at @HPE_HPC.
Bill Mannel is vice president and general manager of HPC & AI Segment Solutions at Hewlett Packard Enterprise.