In its efforts to promote green computing, BNP Paribas Corporate and Investment Banking has the implemented a new GPU-based architecture that the firm says simultaneously reduces electricity consumption and accelerates calculation times. A significant portion of the calculations, about one teraflop (one thousand billion calculations per second), performed for Global Equities and Commodity Derivatives is transferred to a graphical processing unit based platform. This allows a 100-fold increase in the amount of calculation achieved per watt. This new platform, containing two Nvidia Tesla S1070 modules consuming 2kW, will replace more than 500 traditional CPU cores consuming 25 kW.
This initiative involved a re-architecture of the calibration software and its port. The benefits include overall reduction of electrical consumption by a factor of 190, overall reduction of response times by a factor of 15, improved legibility of market prices by traders, reduction in total cost of operation and reduction in environmental impact.
Within BNP Paribas, one of the missions of Global Equities & Commodity Derivatives is to manage portfolios of complex financial instruments. The GECD quantitative research team simulates the behavior of financial markets and is constantly striving to improve the performance of the software. With this goal in mind, the team decided to take advantage of NVIDIA Tesla solutions for the calibration of some of our models' parameters, a process which involves their estimation based on raw financial instrument prices using numerical optimization algorithms.
"We are extremely pleased with these performances, which significantly exceed our initial expectations. We hope to transfer more calculations to the GPU architecture in the near future," says Stphane Tyc, Head of GECD Quantitative Research, BNP Paribas CIB.