The concept of GPGPU originally supposed to use the computational resources of GPUs for financial calculations, we often heard of successful applications of such systems to solve scientific or engineering problems. Just last week, NVIDIA announced the introduction of serious breakthrough in accelerator Tesla in the sector of financial analysis.
Investment bank JP Morgan took a course on modernization of its systems involved in risk analysis. CPUs give way to the accelerator Tesla M2070, based on NVIDIA GPUs with support for CUDA. Now, more than half of all calculations related to the definition of financial risk to JP Morgan is running on hybrid systems that combine CPUs and accelerators Tesla. Speed of payment was increased by 40 times, and planned cost savings as high as 80%. Such an upgrade allowed the bank a three-year plan to reduce costs for risk calculation by 75%.
The calculation of risk can now take minutes rather than hours. It gives financial analysts with a significant competitive advantage in speed to make investment decisions. The loading degree of graphics processors in the respective applications does not fall below 70% for 24 hours - this means that the systems are working efficiently and quickly pay back themselves. JP Morgan also recently managed to get the industry award for an innovative method of risk calculation.