Leverage Extreme Performance with GPU Acceleration
March 7, 2018 Bill Mannel, VP & GM, HPC Segment Solutions and Apollo Servers, Data Center Infrastructure Group, Hewlett Packard Enterprise
Hewlett Packard Enterprise (HPE) and NVIDIA have partnered to accelerate innovation, combining the extreme compute capabilities of high performance computing (HPC) with the groundbreaking processing power of NVIDIA GPUs.
In this fast-paced digital climate, traditional CPU technology is no longer sufficient to support growing data centers. Many enterprises are struggling to keep pace with escalating compute and graphics requirements, particularly as computational models become larger and more complex. NVIDIA GPU accelerators for HPC seamlessly integrate with HPE servers to achieve greater speed, optimal power efficiency, and dramatically higher application performance than CPUs. High-end data centers rely on these high performance solutions to satisfy evolving industry demands, and to create a flexible and agile architecture.
The three types of GPU acceleration
As enterprises endeavor to solve the world’s greatest problems, NVIDIA GPU-accelerated computing will become increasingly vital to support a range of compute-intensive tasks—from commercial applications to deep learning workloads. HPE and NVIDIA offer three categories of acceleration to enhance data center operations and drive innovation: computational, graphics, and virtualization.
Computational acceleration plays a tremendous role in data-heavy processes such as seismic processing, signal and image processing, searching and encryption, financial computing, and computational physics. NVIDIA GPU performance is rapidly outpacing the limited performance of CPUs, leveraging more cores to turbocharge HPC workloads. As a result, enterprises operating on clustered HPC systems can use dense cluster GPUs—with the NVIDIA CUDA Toolkit—to offload code from CPUs with 8-16 cores to GPUs with over 2,000 cores.
Computational acceleration enhances data centers with extreme scalability and performance to tackle their most challenging workloads. For example, enterprises will have the capacity to unlock cognitive computing capabilities, utilizing massive amounts of data to train artificial intelligence (AI) and deep learning algorithms. Once these models are trained, enterprises can deploy smaller models to servers at the edge, where they can analyze data and extract real-time intelligence.
Graphics acceleration is designed to improve engineering and design applications. Predominately used in scientific and manufacturing design fields, enterprises rely on graphics acceleration to visualize patterns in data, fuel life sciences research (such as the Living Heart Project), and much more. Backed by robust HPC servers, NVIDIA graphics tools like NVIDIA Quadro are pushing the boundaries of graphics technology. These solutions increase graphics output and modeling capabilities, which allows users to quickly manipulate and test 3D prototypes—and during downtime, they can be adapted for GPU compute.
Virtualized graphics acceleration is a new development in the HPC market. Virtualized graphics hardware is being used to provide a high-quality graphics experience to end users across a network, straight from the data center. This allows multiple users to access the same graphics-enabled device from their own virtual machine. Virtualized solutions increase workplace productivity and collaboration by equipping users to work remotely using the same GPU for both graphics and compute. And without an expensive dedicated workstation, enterprises will get better economics out of their hardware resources, while ensuring that data remains secure in the data center.
HPE offers multiple server families integrated with NVIDIA Virtual GPU software, coupled with NVIDIA Tesla GPUs to enhance virtualized graphics acceleration. The NVIDIA virtual GPU software portfolio includes NVIDIA GRID vPC and vApps for knowledge workers using office productivity apps, and NVIDIA Quadro Virtual Data Center Workstation (Quadro vDWS) for creative and technical professionals. With Quadro vDWS, users benefit from the power of a Quadro workstation in a virtualized environment. Leveraging NVIDIA virtual GPU technology, when users log into their virtual machines, the NVIDIA driver sends a command to the virtual GPU engine, which schedules an action for the physical GPU. The result is then transmitted to the virtual machine within nanoseconds.
Integrating GPU and HPC solutions
According to the TOP500 List for November 2017, roughly 20% of the TOP500 systems have accelerators, and 85% of those are NVIDIA—including TSUBAME 3.0 at the Tokyo Institute of Technology, which utilizes HPE SGI 8600 architecture with NVIDIA Tesla P100 SXM2 for GPU-based deep learning. Over 500 applications have been optimized for HPC, including 10 of the top 10 applications that are optimized for GPUs. These solutions let enterprises run their codes on smaller, GPU-enabled clusters with greater output and efficiency.
Harnessing the massively parallel processing power of NVIDIA GPUs, HPE is empowering enterprises with accelerator-enabled embedded systems, designed for maximum density and performance. A number of HPE server families are integrated with NVIDIA GPU accelerators:
- HPE Apollo Systems
- HPE ProLiant Rack and Tower Systems
- HPE Edgeline Converged IoT Systems
- HPE Mission Critical Systems
- HPE Synergy Blade Systems
HPE and NVIDIA are pioneering a new era of innovation. In addition to powerful HPE servers, enterprises are implementing NVIDIA Tesla technologies to drive computational acceleration for HPC and AI/deep learning. Others are utilizing NVIDIA Quadro to improve graphics acceleration, as well as a combination of HPC solutions, NVIDIA virtual GPU technology, and NVIDIA Tesla to drive virtualized graphics acceleration for remote users.
Data centers can benefit immensely from these holistic solutions:
- Accelerator integration at the factory
- Integrated management of systems/clusters with accelerators
- Extra PCIe available for multiple options
- Fine granularity of control from rack, to chassis, to node
- World-wide support and management services
- Powerful partner ecosystems
Together, HPE and NVIDIA are helping enterprises leverage extreme compute performance. To learn more about these cutting-edge solutions, I invite you to follow me on Twitter at @Bill_Mannel. And for the latest news and updates on HPC and GPU acceleration, you can visit @HPE_HPC, @NVIDIAVirt