Lenovo, the fastest growing HPC TOP500 vendor, has expanded its support for multiple Graphics Processing Units, (GPUs), from Nvidia and also Nvidia’s GRID technology, as demand for greater compute power in the data center continues to grow. Lenovo systems will now be supporting the new NVIDIA® Tesla™ P100, P40, and P4 GPUs for traditional HPC and emerging deep learning applications as well also the Tesla M10 GPU and GRID for VDI workloads.
Deep Learning and General HPC
With customers increasingly looking at application areas like deep learning or artificial intelligence (AI), they require the raw compute power housed in GPU accelerators because ordinary CPUs are not able to handle these workloads efficiently. Currently, Lenovo customers across the spectrum of HPC and enterprise are looking to AI and deep learning as key pillars of their future. The Tesla P100 for PCIe based servers is ideal for both HPC and mixed use case data center deployments. P100 is powered by the revolutionary NVIDIA Pascal™ architecture and is ideal for demanding HPC applications, with up to 4.7 TeraFLOPS of double-precision performance. With this performance, a single P100 node can replace up to 32 traditional CPU nodes. For deep learning training applications, Tesla P100 provides the highest performance.
The Tesla P40 and P4 are also powered by Pascal architecture and specifically designed for inferencing, which allows customers to use established patterns within neural networks to recognize items like images, text or speech in response to user queries. These Pascal-powered GPUs feature specialized inference instructions based on 8-bit (INT8) operations, delivering up to 22 TOPS (tera operations per second) of performance for inferencing use cases, representing 45x faster response than CPUs, and a 4x improvement over GPU solutions launched less than a year ago. Single-precision support in the ultra-efficient, small form-factor Tesla P4 is 5.5 TeraFLOPS.
VDI-enabling graphics-accelerated Virtual Desktops for all
Organizations have been turning to virtual desktop infrastructure (VDI) to give employees more flexible, secure and productive ways to work for the last several years. These workloads are increasingly moving to newer, graphics-rich applications and operating systems, such as Microsoft Office 2016/365, web-browser based SaaS applications and even Windows 10. Lenovo is helping accelerate the transition to these environments by offering high performance, data center VDI solutions that support these increased graphics needs with NVIDIA GRID and the new NVIDIA Tesla M10 GPU in Lenovo’s rack and dense server products.
The Tesla M10 is optimized for high density Virtual PC applications, where maximized user-per-host density is desired. The M10 can support up to 64 users per board, with up to 128 users supported when deployed in a typical server node with two M10s.
Lenovo solutions built on NVIDIA GRID and the Tesla M10 address the growing dependency that the newest applications and operation systems have on intensive, high-quality graphics. Now, even office and mobile workers demand virtual workspace performance that’s as good as a physical PC. With NVIDIA GRID and the Tesla M10, Lenovo is able to deliver these workers the best virtual user experience possible.