Petaflops for Progress Through High Performance Computing

It would take less than a month and a half for the COVID-19 pandemic to spread from central China to India, with the first diagnosis in southwest India in Jan 27, 2020. Like the rest of the world, the next two years would chart a challenging path for our country to come out of the pandemic.

The country’s researchers joined hands to bring the fight to COVID. But to do it, they needed not just centrifuges and test tubes but also, computing power. And this is where, companies like AMD stepped in.

We are a 53-year-old technology company, that develops computer processors, graphics cards and related electronics. As the high-performance computing leader, we deliver technologies to accelerate a full range of data center workloads - from general-purpose computing to technical computing, cloud-native computing, and accelerated computing - providing scientists, engineers, and designers faster insights and more accurate results.

In April 2020, we launched the High-Performance Computing (HPC) Fund to accelerate medical research into COVID-19 treatments. The Fund was an excellent example of how public and private partnerships can help tackle some of the world’s toughest challenges.

Under the HPC Fund, AMD made available computing resource grants to public institutions who then could conduct the much-needed research. We have thus far awarded more than 20 petaflops, or 20 quadrillion floating-point operations per second, through the program. If we consolidate the resources, we will rank 28th among the world’s 25 fastest supercomputers, as of November 2022.

What gives me a great sense of pride is the fact that the largest single on-premise donation under our COVID-19 HPC fund initiative was made in India.

Partnership with CSIR-4PI

AMD formed a committee of HPC experts comprising of technologists and corporate fellows, government-facing teams, domain experts, and those who worked closely in computer labs. The group then drew up a list of organizations across India that could benefit the most from this HPC asset. These assets are physical computing clusters that were to be installed on-site at server facilities.

From a long list of possibilities, six possible candidates emerged as leading contenders. Among them was India’s Council of Scientific & Industrial Research (CSIR), a nationwide network with 79 facilities and a user base of over 8,000 active scientists and supporting personnel.

Already a major AMD customer, the institution quickly rose to the top as a viable hosting partner thanks to its robust server infrastructure and network. CSIR also had the means to host and maintain this large infusion of computing power.

CSIR has been hosting server farms since 1994. In 2020, they were leading most of India’s COVID-19 research. Funding from AMD was going to be the largest donation of computing power for them.

Located in Bengaluru, the institute does major research around data modeling, artificial intelligence, biological genomics and computer simulations. It even has its own data science and supercomputer division (DSSD), along with an academic wing that credentials researchers.

Speaking about the opportunity, Dr. Gopal Krishna Patra, Chief Scientist at the CSIR Fourth Paradigm Institute (CSIR-4PI) told us, “When we saw the request for proposals, we were really excited because CSIR was actively contributing to COVID-19 research,”.

In the fall of 2020, AMD approved the CSIR application. CSIR Director-General Shekhar Mande told The Hindu Business Line at the time that the capabilities would “augment our capacity...to provide world class high-performance computational facilities to the research community.”

Why Computing Power Matters?

Much of the work being done In India is in the trenches of protein syntheses and related genomic simulations. These efforts are traditionally laborious, systematic processes, often done by hand. Computers speed them up exponentially, but the gains are capped by the physical hardware at a given laboratory or available via network connections. There is seldom enough computing power, a problem that was visibly and dramatically evident during the pandemic.

This is where supercomputers become vital. In the time it takes a researcher to conduct a few hundred protein-synthesis processes by hand, a supercomputer can do a few million. In a month, it could process the trillion-plus combinations of protein synthesis. And that is essentially how drug discovery happens.

The Next Step - Creating a User Network

The AMD HPC system at CSIR was not just for the use of the institute, but for any researcher who needed computing power for COVID research in India. To find and qualify users, the institute in partnership with AMD developed a review and approval system. In what became known as the COVID CARE Network (CCN), researchers across India could submit proposals to secure remote access over a high-speed national knowledge network.

There was no requirement that CCN work must be part of CSIR. It was open to all. The only condition was that whatever the research outcome is, it should be made public. At the core was AMD’s philosophy of open-source innovation and data sharing.

As of now, four proposals have been approved for CCN users. These proposals have come from institutes that lack native supercomputing resources.

Dr. Malay Rana is among the beneficiaries. An assistant professor of chemical sciences at the Indian Institute of Science Education and Research Berhampur (IISER), Rana and his team had to run as many simulations as possible for their COVID research but came up short on computing resources.

Speaking to us, he said, “We really struggled to do our research because of computational limitations. The pandemic dashed an earlier plan for IISER to have its own HPC facility, a resource the institution still lacks”.

Before the COVID Care Network, Rana was running simulations off a few dozen desktop computers. “We were struggling. And then in September of 2021, we came upon the news that this huge computational facility was being made available by CSIR and AMD to researchers”.

IISER immediately applied. Withing three months Rana and his team were able to begin using the computing resources and carry out their research at an exponentially faster pace compared to before.

Expanding the scope of HPC Fund

AMD’s global HPC Fund is an ongoing initiative, with the most recent expansion announced in June 2022. The new infusion of computing power will raise total global capacity from the fund by about 70%. Some 6,000 researchers have already tapped into the resources, and dozens of papers have already been published.

AMD has now also broadened the scope of this HPC Fund. Going forward it will cover research into climate change, health care, transportation, big data and more. It will help institutes like CSIR advance the scope to use supercomputing resources for non-COVID research including ongoing research areas like cancer and asthma.

Needless to say, researchers like Dr. Malay are excited. “Our role is to show research to people who are actually synthesizing the drug molecules and performing the clinical trials,” Dr. Rana said. “But without the computational support, the experimental research community will have a hard time discovering anything within a given period of time. We are really looking excited about the news of additional resources for our research.”