Petrobras assembles Pegasus, its largest supercomputer, with a capacity of 150,000 laptops

RIO – When your job is to find Petroleum hundreds of miles off the coast and miles of depth, nothing better than calling on mythological beings for help. After Phoenix, Atlas and Dragon, the Petrobras called upon the Pegasus supercomputer to assist its geophysicists in the difficult task of finding the right place to drill exploratory wells, or identify the existence of oil at the bottom of the ocean. After all, a puncture in the wrong place can cost the company $100 million. With better-defined seismic images, Pegasus will reduce geological and operational risks, as well as the time between the discovery of a field and its entry into production.

Pegaso will be the state-owned company’s fourth and largest supercomputer and the seventh in Brazil, a country well placed in the world ranking of supercomputers according to the professor at Coppe/UFRJ Álvaro Coutinho, director of the university’s advanced core of high-performance computers.

“Brazil is the 11th country in the world in terms of number of supercomputers, three of which belong to Petrobras, which shows that it has a high technological level. We are very well placed and, with Pegasus, we will be among the top 10”, said Coutinho.

Assembly began in July, in Vargem Grande, Rio de Janeiro, and by the end of the year the new 30-ton supercomputer will start operating. When up and running, it will have the same capacity as 150,000 laptops or 6 million cell phones. It took 32 trucks to carry all the pieces, which, aligned, correspond to the height of a 10-story building (35 meters).

Pégaso joins Fênix, Atlas and Dragão, three supercomputers installed in the last three years and which raised the state-owned company’s processing capacity from 10.2 petaflops in 2019 to 42 petaflops in 2021. Each petaflop – acronym for Floating- point Operations Per Second – performs 1 quadrillion calculations per second.

Petrobras’ master consultant for 35 years in the technology area, Luiz Rodolpho Monnerat says that supercomputers began to arrive at the company in 2007, after the discovery of pre-salt, and took a few years to reach the current 42 petaflops. Petrobras was a pioneer in Latin America in the use of graphics cards to do seismic processing and one of the first to also use high performance computers.

With Pegaso, the oil company will double its capacity and close the year with 80 petaflops, adding other small computing clusters, one of them nicknamed Tatu. Alone, the new equipment has 62 petaflops, doing in seconds, for example, what a laptop could take weeks.

“Since the 1970s, Petrobras has been working with high-performance computers, but you can’t compare it to Pegasus. As processors get faster, you can solve more challenging numerical problems and bring more savings to the company,” explains Monnerat.

• More precision

Pegasus makes it possible to generate more accurate seismic images of the subsurface of areas where Petrobras wants to produce or discover oil, such as the Equatorial Margin, a new frontier that the state company is about to open. “You have less chance of drilling a dry well 100 meters from the right place, for example”, explains Luiz Rodolpho Monnerat, Petrobras’ master consultant for 35 years in the technology area.

• Economy

“With a better image, instead of putting eight wells in the field, you make six, because you can see the best places to drill”, says consultant Luiz Rodolpho Monnerat, noting that, with this, Petrobras saves a few million dollars.

• Viability

According to Monnerat, 10 years ago, a computer with this capacity cost at least US$ 3 billion and today the company pays US$ 300 million for using it for five years, when it should become obsolete. “There have always been supercomputers, but 10 years ago it was impossible to build a Pegasus, the economic return would not be the same. We invest in this because we expect a greater return than the investment”, explains the consultant.

• Capacity

With Pegasus, the oil company will double its capacity and close the year with 80 petaflops (a unit of performance, especially in the processing of calculations), adding other small clusters (sets) of computing. Alone, the new equipment has 62 petaflops, doing in seconds, for example, what a laptop could take weeks.