India’s supercomputers are falling behind, despite growth in overall power

Despite growth in India’s overall supercomputing capabilities, the country’s overall supercomputing output has been dropping in the list of the world’s most powerful machines. From having two of the top 100 supercomputers in the world in 2020, the country now has none, and it has only three supercomputers in the Top500, a global ranking service for the fastest supercomputers in the world.

Supercomputers can accelerate calculations like this by providing exponentially larger computing power than regular computers. In May 2020, researchers at US Oak Ridge National Lab ran thousands of simulations to find drug compounds that could stop the covid-19 virus from infecting host cells. They completed the simulations in just 2-3 days using IBM’s Summit supercomputer. 

“Supercomputers can be utilized for a large number of things in certain areas like chemistry formulations, protein folding, biomedicine. It can also be used in space, for satellite placements, which are very, very difficult to solve even with classical computers,” said Nikhil Malhotra, chief innovation officer at Tech Mahindra.

Countries around the world, including India, have been ramping up their supercomputing capabilities over the past few years. In 2022, the Ministry of Electronics and Information Technology (MeitY) has installed four supercomputers at various institutions in the country. 15 supercomputers, with an aggregate compute capacity of 24 petaflops, have been installed in the country since 2015 under the National Supercomputing Mission (NSM).

According to experts, building local supercomputing capabilities will strengthen high-quality research in biomedicine, space tech, and climate, which require high compute power. However, they also pointed out that the available compute power of Indian supercomputers is considerably lower than their global counterparts.

India’s first supercomputer, PARAM 8000, was set up in 1990. Param Siddhi, which offers a peak performance of 5.27 PFlops, had the distinction of being ranked as the 63rd fastest supercomputer in the world when it was launched in November 2020, according to the Top 500 global ranking service. It has since slipped to the 111th spot.

Similarly, the Indian Institute of Tropical Meteorology’s (IITM) Pratyush supercomputer, which used to be in the top 100, is now ranked 132. Overall, India has three supercomputers in the Top 500’s overall list. China and the US account for nearly two-thirds of the top 500 supercomputers in the world, with 173 and 128 of the 500 most powerful supercomputers in the world, respectively.

The computing power of a supercomputer is measured in floating-point operations per second, or FLOPS. One PetaFlops is equal to 1,000,000,000,000,000 (one quadrillion) FLOPS, or one thousand TeraFlops. The world’s fastest supercomputer, Frontier, located at the Oak Ridge National Laboratory, offers a peak performance of 1,685 PFlops.

“The baseline behind every supercomputer is faster performance. The world is moving to ExaFlops and we are still on PetaFlops. We are always two years behind. If we continue to play catch up we will always remain behind,” said Preeti Syal, a tech policy expert, who used to work with Niti Aayog. 

Suryachandra Rao, project director at IITM also noted that leaders like China and USA are looking at exascale level computing and that is the next milestone that India also has to reach. However, he also pointed out that India faces a dual challenge of expanding the capability and also making such resources for more researchers and institutes. “It is a step in the right direction,” he said.

“At this moment the idea is to enable more people to use such resources. Ten years back, we used to have 1 or 2 systems. Now we are setting up at different locations. This will build capacities and allow more users to access it,” he added.

Rao said that earlier weather and climate forecasts were the only priority. Now there are many other applications for supercomputing. 

But in addition to having supercomputers, having the latest hardware is equally important. Syal noted that how India is making these supercomputers or the components used in them is also important to take their performance of supercomputers to the next level.

Building more cases is the next challenge for the NSM. Syal pointed out that currently, the use cases are limited. “Globally, most of the usage comes from industry. What we are doing instead is letting research institutions drive it. They will never be able to think in terms of commercialization. These research institutions should also be linked directly to big industry players,” she added.

Budget is another challenge. Rao noted that his institute is working on increasing the capacity of Pratyush. Though he said that their requirement is at least 150 PFlops they will be able to get somewhere around 10-20 PFlops due to budget constraints.

C-DAC didn’t respond to Mint’s email query till the time of writing.