Optimising Monte Carlo option pricing using GPUs

Ryan Saunders
Department of Computer Science, University of Cape Town
University of Cape Town, 2013


   title={Optimising Monte Carlo option pricing using GPUs},

   author={SAUNDERS, RYAN},



Download Download (PDF)   View View   Source Source   



Computer modelling has been used for a number of years already to aid financial institutions in making business decisions. One such decision that financial firms are often faced with involves setting fair prices for financial options. Since the process of option pricing can be computationally expensive, methods of optimising it are sought after. One popular method is to use parallel programming architectures to split the problem among many processing cores. In this regard, GPU-based solutions are known to have massive performance benefits, when compared to multi-threaded CPU-based solutions. In this paper we present a performance comparison between using a quad-core CPU and a GPU to price Asian options. We use an embarrassingly parallel stochastic solution, Monte Carlo simulation, to derive both the option prices and Greeks, providing useful data for financial experts and allowing for the possibility of producing explanatory visualisations. We also observe that while the use of quasi-random sequences often improve the Monte Carlo convergence rate, it does not work well for large volatilities. We find that, for our task, a program utilising a GeForce GTX780 can outperform an optimised multi-threaded program running on an Intel Core i5-650 CPU by a factor of at least 49.3x.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2020 hgpu.org

All rights belong to the respective authors

Contact us: