What is common to Gravity, The Hobbit and Iron Man 3? They were all nominated for the Oscar this year for their hi-tech visual effects from design studios such as Frameworks, IL&M and Pixomondo. Again, what is common among these?

They all use workstations with NVIDIA graphic processors or GPUs.

At a recent NVIDIA technology conference at San Jose, as I was finding my way around a host of geeks and nerds, here was Sumit Gupta from NVIDIA who was willing to give simple answers to complex issues. Indian-born Sumit Gupta, with an IIT degree followed by a Ph.D from the University of California, Irvine, and with 20 years in software and chip technologies, leads the Tesla GPU Accelerated Computing Business group at NVIDIA.

Tesla is popular among users of high-performance computing machines used in data analytics, technical and scientific computing. Gupta has to his credit one book, several chapters in books, one patent and several technical papers.

Excerpts from the interview:

How do you foresee the use of GPUs for analysing big data?

GPUs are becoming a central part of machine learning and artificial intelligence methods. Machine learning is a significant solution to big data analytics and GPUs play a key role in this. We have had several achievements in this area. For example, you can teach the computer to recognise patterns. We can take Twitter as an example. You can tell the computer, “Here is all the history of Twitter and you analyse it.”

The computer might say that there is a weather storm coming, because the last few times there was a storm, Twitter noise (conversations) had become loud.

Or, in the case of the Arab uprising, perhaps, had Twitter had been analysed correctly, the Arab Spring could have been predicted. GPUs help make such analysis on social networking platforms fast and efficiently.

How do you use the Tesla GPU for big data analytics?

Fundamentally, we use the Tesla GPU for analysis whenever there is a big ‘computational’ task in a server — big data analytics or database analytics.

For instance, it can be used to analyse call records for a company, say, Reliance. Reliance has a lot of call records and it would like to find out if customers who exhibit a certain behaviour are likely to buy a certain service, because such has been experience. So, it can proactively offer this service.

The market for GPUs is quite diverse. The traditional (easiest) market for GPUs is high-performance computing, which includes government and research laboratories.

ISRO and ANURAG (Advanced Numerical Research and Analysis Group) are some of the Indian institutions using GPUs in the government sector. Also, academic centres such as IITs and IISc (Indian Institute of Science) use GPUs for research — analysis of medical imaging, genome sequencing, cancer detection, weather prediction and any area that needs scientific computing.

Even the ONGC technologists are doing imaging under the earth to find where oil is; computing is done even to improve weather forecasts so that tsunami and monsoon can be predicted more accurately. We enable Indian scientists and technologists to do things they couldn’t do before.

What role do GPUs play in the detection of diseases, say AIDS?

People are trying to figure out how to attack the HIV virus. So, the way drugs work is that they connect to the virus (called docking) and change the property of the virus. The drug plugs a hole and deactivates the virus. Scientists take all known drugs and chemicals on the database to match them with the appropriate virus. However, done in traditional ways, this may take years using CPUs.

The GPUs help to do this faster. In effect, we are accelerating the search for new medicines by improving the computing power. In one particular case, we had molecular modelling done on the computer and analysed the interactions using 300+ GPUs. This kind of scientific research is being done not just for HIV but for many other diseases as well.

The writer was in San Jose at the invitation of NVIDIA to attend GPU Technology Conference 2014

Related Topics