Reval Reduces Time to Price OTC Derivatives

"...One of the tools NVIDIA has introduced, and a key driver behind the adoption of GPUs in financial markets, is a programming environment called CUDA, which was released two years ago. It is this CUDA architecture inside the chip (as well as a CUDA toolkit and a C-compiler) that enables software developers to tap into the parallel architecture of the GPU using the C programming language.

Based in Austin, Texas, SciComp, a developer of derivatives software, began using NVIDIA's CUDA technology in 2007 so that it could parallelize its code. The vendor's SciFinance product "actually generates the source code for the pricing and risk models," explains Curt Randall, EVP and head of financial engineering as well as a founder of SciComp.

The thousands of lines of C or C++ code generated by SciFinance can help a quant produce a new pricing model in a matter of hours, says Randall. "The compiler and the language make the hardware accessible to general programming as opposed to graphics programming," says Randall, who contends that the NVIDIA hardware can enable the code to run 30 to 50 times faster than standard PCs. "Programmers don't have to understand the nuances of how to parallelize code. All of that is done for them automatically."

This article appeared in Wall Street and Technology.