October 2016
by Alan Griffiths
@cambashi_alan
LinkedIn

NVIDIA is a company that most people associate with their graphics card, but at its GTC event, I discovered it is much more than that. Jen-Hsun Huang’s keynote speech highlighted three particular areas of interest.

1. Supercomputing and AI

NVIDIA is much more than ‘a company that makes graphic accelerator cards’; it has become a major player in supercomputing, neural networks and artificial intelligence.

Figure 1: The NVIDIA DGX-1 stand

This was illustrated by the announcement of the DGX-1 AI supercomputer for deep learning which has been adopted by several AI institutes. These include the following:

DFKI (German Artificial Intelligence Institution)

DFKI (German Artificial Intelligence Institution) bills itself as ‘the world’s largest AI research centre’.  On September 28th NVIDIA and DFKI announced a collaboration KI that includes research funding over four years, as well as the donation of an NVIDIA DGX-1 deep learning supercomputer.  Dr Damian Borth, Director of the DFKI Deep Learning Competence Center said: “The strengthening of our co-operation with NVIDIA will further accelerate the considerable advances DFKI has made in the domains of self-driving cars, multimedia opinion mining, emergency response and Industry 4.0.  It complements the ambitions of the Deep Learning Competence Center to enhance basic research and industrial knowledge transfer in the field of AI.”

Figure 2: NVIDIA’s AI announcement

ROMEO high performance computing centre

In France, DGX-1 will help the team at the University of Reims Champagne-Ardenne’s ROMEO high performance computing centre use deep learning to help prevent or cure diseases affecting grapevines.

“Since the 1980s I have been working on building an AI smarter than myself so I can retire. The methods we have developed on the way to this goal are now heavily used by the world’s most valuable public companies.” – Professor Jürgen Schmidhuber, Scientific Director of IDSIA

IDSIA:  Switzerland’s leading AI laboratory.

IDSIA focuses on machine learning (deep neural networks, reinforcement learning), operations research, data mining and robotics.  On September 28th NVIDIA and IDSIA announced a collaboration KI that includes research funding over four years, as well as the donation of an NVIDIA DGX-1 deep learning supercomputer.  Professor Jürgen Schmidhuber, Scientific Director of IDSIA said: “We are delighted to extend our interaction with NVIDIA through this initiative. Since the 1980s I have been working on building an AI smarter than myself so I can retire. The methods we have developed on the way to this goal are now heavily used by the world’s most valuable public companies. But much remains to be done, and NVIDIA’s support will help us continue to push the limits.”

As well as AI institutes, NVIDIA is breaking into industry’s use of AI:

SAP

SAP is now using DGX-1 AI supercomputers at its German and Israeli offices, where its teams are building machine learning solutions for SAP’s 320,000 customers.

2. High Performance Computing

By positioning its ‘platform’ GPU products for computer hardware manufacturers and cloud datacentre providers, NVIDIA has made its technology central to High Performance Computing.

This was shown by IBM’s announcement that its new ‘Minsky’ supercomputers, based on their Power 8 chip, will use Tesla P100 GPUs.

Also, Microsoft’s Azure N-Series virtual machines will use GPU visualization and GPU compute infrastructure enabled by the NVIDIA Tesla M60 platform with GRID and Tesla K80 GPU accelerators.

3. Visionary technology – autonomous vehicles and medical research

Around 20% of NVIDIA’s annual investment goes into developing ‘visionary’ technology to support vertical markets such as medical research and autonomous vehicle solutions.

“For the current application of our AI technology in the bioscience space, it will mean that new drug discoveries can be made faster and more efficiently than ever before” – BenevolentAI co-founder and Director Ken Mulvany

BenevolentAI

In the U.K., BenevolentAI will use DGX-1 as part of its effort to accelerate drug discovery by using deep natural language processing, machine learning and artificial intelligence to formulate new, usable knowledge from complex scientific information.  BenevolentAI co-founder and Director Ken Mulvany said: “For the current application of our AI technology in the bioscience space, it will mean that new drug discoveries can be made faster and more efficiently than ever before”.

Self-driving cars

NVIDIA are working closely with automotive companies like Audi and Volvo, both of whom had autonomous vehicles on display.  They also announced at the event that they are partnering with Tom Tom (based in Amsterdam near where the event was being held) to develop artificial intelligence to create a cloud-to-car mapping system for self-driving cars.

By providing the required hardware, such as the Drive PX 2 and Xavier AI supercomputer for in-car deployment, along with software platforms and SDKs, NVIDIA is positioning itself to benefit from the estimated $10 trillion transportation market.

The Xavier SoC (System on Chip) will provide an on-board inference engine for autonomous vehicles.  The Xavier SoC manages 20 trillion operations per second, while only using 20 watts of power.  Xavier is designed to be used as an AI brain for use in self-driving cars specifically.  Nvidia CEO Jen-Hsun Huang told attendees at the conference that it’s “the greatest SoC endeavor” he’s ever encountered.  Because it’s used in cars, Xavier was designed to meet the ISO 26262 functional safety spec, which is an international standard that sets expectations for electronics used in cars designed for road use.  The SoC uses a 16nm manufacturing process, and just one can replace Nvidia’s current DRIVE PX 2 in-car computer, including a configuration of said component that includes two mobile SoCs and two discrete GPUs, while also using less power.  Xavier is intended for use by carmakers, suppliers, research organizations and startups looking to field and test their own self-driving cars. You won’t see it in any cars in the near future, however — Nvidia says it will start shipping the first samples in the fourth quarter of next year (2017).

Video: Volvo’s self-driving car demo – powered by NVIDIA

NVIDIA, at GTC, demonstrated the advanced state of autonomous driving using massively parallel supercomputers based on the DGX1 for ‘training’, and the in-car ‘SoC’ (system on chip) devices mentioned above to provide the inference needed for rapid reaction in real-time.

Conclusion

NVIDIA is demonstrating its capabilities in a wide range of high-performance computing areas and is quietly building a position in Machine Learning, Artificial Intelligence and Autonomous Vehicles. All of those areas are still at the exciting stage where anything is possible. The next few years will determine who are the winners and losers. NVIDIA will hope to be amongst the winners.

Your views

If you were at NVIDIA’s conference, let us know what you thought.