NVIDIA's New GV100 Deep Learning GPU with Volta Architecture

Andrew Wheeler posted 2018-03-27

At today's GPU Technology Conference, NVIDIA founder and CEO Jensen Huang revealed a remarkable new Volta-based Quadro GV100, and described how it transforms the workstation with real-time ray tracing and deep learning. Huang displayed real-time ray tracing during a dazzling Star Wars demo created by NVIDIA, Industrial Light & Magic and Unity. All the complex and intricate ray tracing of light in this demo was done in real-time. 


The Quadro GV100 conjoined with the Quadro vDWS for the data center, are going to have a big impact on the capabilities of industrial sectors like automotive, architecture, engineering, entertainment and healthcare. These industries are creating an up swell in demand for deep learning-based research and development, create photorealistic VR, and create more advanced deep learning-enhanced applications. 

Capabilities provided by these new GPUs will empower designers, engineers and scientists to leverage deep learning, rendering and simulation to create better and more targeted products. Generative design will benefit from these capabilities. Simulations will become faster and more capable of running more complex scenarios.

A first look at the new Quadro GV100, which has 7.4 TFLOPS double-precision, 14.8 TFLOPS single-precision and 118.5 TFLOPS deep learning performance, and 32GB of high-bandwidth memory capacity. Two GV100 cards can be combined using NVIDIA NVLink interconnect technology to scale memory and performance, creating a massive visual computing solution in a single workstation chassis. (Image courtesy of NVIDIA.)

Other benefits of the GV100 include:

  1. Implementing deep learning developments quickly and easily – Access a huge catalog of GPU-optimized software tools from the NVIDIA GPU Cloud container registry with GV100 or other high-end Quadro GPUs for deep learning and high performance computing on any workstation.
  2. Turning your desktop workstation into a deep learning and inferencing superhero – Scale up your machine with two GV100s for up to 64GB of HBM2 memory connected by NVLink. Speed up your deep learning training and inferencing explorations for creating AI applications.
  3. Deep-learning based rendering and ray tracing  – Deep learning-accelerated denoising for ray tracing is the new linchpin for fast and robust ray tracing on your design workstation.
  4. Scale up memory and perform complex 3D simulations with double-precision – Hyper-quick double-precision conjoined with an ability to scale memory up to 64GB boosts solver performance in computer-aided engineering workflows.
  5. Design, create and collaborate in VR – Designers now have support for advanced VR features and can create physics-based VR platforms like NVIDIA Holodeck for design reviews of large-file products and the ability to immerse users in photorealistic scenes and products at scale.

Quadro vDWS

Quadro vDWS allows users to do the following things:

  1. Run interactive, real-time simulations like ANSYS Discovery Live
  2. Reduce the rendering time of photorealistic images by 80 percent compared to previous technologies
  3. Use AI-enhanced applications for more visual fluidity and interactivity in design workflows
  4. Work from any connected device anywhere in the world while data remains secured, great for distributed teams

Availability of Quadro GV100 and Quadro vDWS

On stage, Huang mentioned that the two Quadro GV100 GPUs coupled together with NVLINK cost $68,000, which sounds like a lot unless you are comparing the cost to a supercomputer used by major CGI firms like ILM. The Quadro vDWS is available now for 120 systems from 33 vendors. The NVIDIA Quadro GV100 is available now on nvidia.com and starting in April from leading workstation OEMs, including Dell, HP, Lenovo and Fujitsu, and authorized distribution partners, including PNY Technologies in North America and Europe, ELSA/Ryoyo in Japan and Leadtek in Asia Pacific.