In a candid conversation with Vishal Dhupar, Managing Director, South Asia, NVIDIA, Dataquest tried to understand the company’s intent behind bringing GTC to India and what it would mean to the customers and partners. Here are the excerpts from the conversation:
In what ways is the effective leadership at shaping up the company’s future?
Jen-Hsun Huang, NVIDIA CEO, is one of those very rare leaders who have hands-on technical as well as business acumen. During my lifetime, I have worked with multiple companies and seen leadership from different aspects. But seldom have I seen someone like Jen-Hsun Huang who can communicate with the engineer about the design of architecture and with the business manager about the Wall Street. He has an accomplished personality and we are fortunate to have a founder like him.
There was a time when NVIDIA was hardly known for business beyond graphic cards and lacked communication in the outside world. But today, the company is far different from the past as it is investing into AI and VR. What has caused this transformation?
That’s right. Beyond a certain point you cannot talk about chips. We have a new deep learning system; DGX-1 during the GTC Conference in India. Knowing our business model, the industry experts have begun to wonder as to what we will do with the DGX systems. There are questions if we are trying to be an original equipment manufacturer (OEM). But let me tell you that Jen-Hsun Huang set up the company in 1993 with a vision to address the challenges around visual computing to empower it. Indeed, visual computing is just one area but very critical.
Jen-Hsun Huang once said, “We are in business to solve problems which are under the umbrella of visual computing.” His endeavor has been to figure out the biggest application for visual computing which is still unsolved on the x86 Windows platform, and that area was gaming. Today, interestingly gaming industry is two times the size of the movie industry, and three times the size of the music industry. All in all, the gaming scores over a $100 billion market.
Keeping in mind the excessive professional and personal usage of PC, Jen-Hsun Huang planned to use it for entertainment. The concept triggered him to build a processor on which the company could actually write a lot of algorithms to develop games.
What makes NVIDIA to increase its focus on deep learning and virtual reality?
Observing the constant developments in the deep learning space, NVIDIA endeavors to solve a lot of problems through innovative products. /In the month of April this year Jen-Hsun Huang announced that the company has a parallel processor which can solve the problems related to AI and deep learning. This is the reason that we launched DGX-1 which is a deep learning graphic accelerator system. It basically is an equivalent of 250 computers into one 3U box. This single box offers a strength of 250 high-end CPU servers at once since the system uses two CPUs and eight GPUs. DGX-1 is also capable of helping companies in the VR space.
In most of my conversations I have had with NVIDIA people, I’ve seen most of your people do not talk like a salesman but like an engineer which is very rare.
This is because we as a consumer-oriented company are trying to figure out, how technology in a use-case can be amplified. It is our effort that customers know about the technology which they are investing in. They need to know that there is something there in it and it allows them to weigh other options. If they feel good about their purchase, they become referrals.
Awareness does not happen overnight. Our teams are focussed on creating awareness and consideration among customers. The partners can come together and make a sale. And that’s the slight distinction between all other companies and what we are doing at NVIDIA.
Since NVIDIA is conducting the GPU Technology Conference in Mumbai, India this year, what is the primary objective of the conference? What sort of potential start-ups are expected to be a part of the conference?
The primary objective of the conference is to bring the brightest minds in the country together under one roof and understand how deep learning can solve many of the computing challenges faced on a regular basis in the industry. One of the biggest reasons to focus on potential start-ups is that our country is really young and our challenge is to create jobs. And, with this entrepreneurial spirit we can solve some of the unique problems. Surely, NVIDIA can add value in this pursuit.
What is the criteria and benefits for start-ups to participate in the GTCx India?
Start-ups can showcase their idea as well as technology through our inception program. The Inception Program creates a platform where start-ups will meet people who can join hands with them. If somebody wants to do a visual discovery in the area of retail but is small and young right now, their focus could be on one aspect of visual discovery. This program will connect them with professionals who are working in the same field and can complement their idea through partnership or otherwise. Inception Program is our vehicle to bring people and business together.
What kind of message do you want to convey to the partners at GPU Technology Conference?
The problems that we are trying to solve during the conference require enormous amount of throughput computing which our architecture is capable of providing. Secondly, there is win-win situation for everyone in this. If we can all come together, we can serve all our commercial aspirations together. Finally, we get noticed at an international fore. The investments will become easier and we can all grow together. All in all, it is a platform for taking all of us a level up in business.
In what ways is the data center space evolving? At the same time, in what ways do you emerge as enablers for DCs based on x86?
In the data centre market, there’s a transformation going on. Today, data centers are capable of solving hyper-scale problems simultaneously. Additionally, there is a challenge to handle enormous data that is getting generated either through high-performance computing or hyper scale. How do you get insight into that information and how to get that knowledge which is at the AI front? Today’s and tomorrow’s data centre should be versatile to support all these needs.
Let’s take the example of Google Brain which is a hyper scale platform and to the level of AI. In 2010, they decided to feed millions of videos to the computer and see what it comes out with. And what it came out was identification of cats. This happened because Google’s data centre utilized 2000 CPUs to perform this. Through multiple iterations, their computer could identify the common pattern of a cat. Stanford took over that project in 2011 to use GPUs as a combination. They used three GPUs and achieved the same result. What does it really mean from an Indian context? You can reduce the size of your data centres. If you reduce the size of your data centres, it means lesser electricity, lesser management. You can increase the velocity for that same investment to do the work to benefit on a smaller investment with a smaller operational cost. And the classical example is from a DGX perspective. You can take 250 base servers on CPUs and reduce that to 2 CPUs and 8GPUs in a 3U form factor – delivering the power of 250 systems at 3kW supporting 7TB of storage and 7TB of memory into that whole thing. Now you can put your whole thing in a system below your desktop which means lesser electricity, faster usage and better build-up. And that’s what the need of the hour is.
When we are talking to the world at NVIDIA GTC in Mumbai, the software programmers need to be the best in all computing models and you still need to provide an affordable infrastructure. And that’s what GTC plans to bring and demonstrate.
By: Onkar Sharma