Quanta Expands GPU Server Lineup to Support NVIDIA P4, P40 Inferencing Accelerators

by Melody Tu

NVIDIA’s reach across the AI ecosystem grew even broader this week with the addition of new servers for inferencing from Quanta. It’s yet another example of how industry leaders are supporting our end-to-end artificial intelligence infrastructure.

Quanta – one of the world’s largest enterprise server vendors – has expanded its server lineup to feature GPUs based on the NVIDIA Pascal architecture, which are purpose-built to be the engine of computers that learn, see and simulate our world.

“Quanta has been working with us on GPU server since the very beginning,” said NVIDIA co-founder and CEO Jen-Hsun Huang, citing Quanta’s work with NVIDIA on DGX-1, the world’s first AI supercomputer in a box; Facebook’s Big Sur inferencing server; and a slim 1U server featuring our candy-bar sized Tesla P4.

“It doesn’t matter what kind of data center you want to build, our partnership with Quanta and QCT (Quanta Cloud Technologies) will make it happen for you,” Huang added.

Huang was speaking at GTC Taiwan, the second in a series of eight regional GPU Technology Conferences being held this fall on four continents. Some 2,000 developers, researchers and executives attended the event – a similar number to those at last week’s GTC China in Beijing.

The event also included:

  • Three dozen breakout sessions on topics such as deep learning/AI, professional visualization, high performance computing, virtualization, embedded computing and VR.
  • A series of Deep Learning Institute hands-on labs to train developers on deep learning techniques.
  • A robotics challenge featuring teams from 10 Taiwan universities.
  • An Emerging Companies Summit in which nine startups described how they’re using GPUs for AI.

“NVIDIA consistently delivers high-performance solutions for our customers’ most demanding, low-latency application workloads,” said Mike Yang, president of Quanta Cloud Technology. “As apps leveraging machine learning, neural nets, AI and big data become more critical to business success, our servers equipped with NVIDIA GPUs will outpace the performance advanced users need.”

Deep Learning Everywhere

Behind the scenes, NVIDIA GPUs are being used to develop powerful new kinds of AI at some of the world’s top research labs, deploy AI-powered services to hundreds of millions of people from hyperscale data centers, and power a new generation of autonomous devices.

Since 2014, the number of GPU developers has tripled to 400,000, and the number of developers harnessing GPUs for deep learning has grown 25x to 55,000.

Wide Array of Options

To support this fast-growing ecosystem, our network of the world’s leading server makers is committed to building a wide range of GPU-dedicated servers.

These systems range from slim 1U systems built around NVIDIA Tesla P4 GPUs — which offer 40x the energy efficiency of CPU-only systems — to burly 4U systems incorporating as many as eight NVIDIA Tesla P100 or NVIDIA Tesla P40 GPUs.

Quanta will be the first to offer systems built around our NVLink interconnect technology, which allow processors — CPUs and GPUs — to exchange data five to 12 times faster than PCI-e (see “What Is NVLink?”).

Quanta and NVIDIA will bring deep learning to everyone from cutting-edge research teams, to hyperscale data centers, to enterprises using deep learning to tackle their toughest business problems.