Advanced Computing in the Age of AI | Thursday, March 28, 2024

Cloud-Based GPU Accelerators Crunch Financials 

Graphics accelerators have been used over the last several years as a complement to or replacement for standard CPUs. GPU leaders like Nvidia have pushed the trend, arguing that graphics processors have advanced to the point where they can also be used for high-end computing tasks.

Now, Nvidia and a partner are targeting specific markets like hedge funds and other financial institutions with a cloud-based, GPU-accelerated processing platform. Nvidia partner Elsen Inc. says it is currently testing the relative merits of in-house versus cloud-based systems for applications like back testing trading strategies or running financial simulations.

Elsen, the Boston-based financial technology vendor, said this week it is using Nvidia's Tesla Accelerated Computing Platform, which includes GPUs and enabling software, to help it crunch numbers and run simulations.

Leveraging the Tesla accelerator, Elsen claims its computing engine runs financial algorithms an estimated 600 times faster than comparable CPU-based systems. It also said the GPU approach reduced the number of servers needed while reducing other operating costs.

The Nvidia-power platform is being used to run complex time-series calculations like Monte Carlo simulations and other algorithms used in the financial services industry for back testing, pricing and risk management. Monte Carlo simulations are used to determine all possible outcomes of a decision to gauge the impact of risk.

Cloud-based GPU accelerators are gaining traction in the financial community because their thousands of high-end processing cores excel at parallel calculations. That translates to speed, which is what traders require to run more complex models with finer calculations, scenarios and sensitivities, Elsen noted.

In November, Nvidia rolled out a new twin GPU Tesla card dubbed K80 that crams more floating-point performance into servers. At the time, the chipmaker described the Tesla K80 accelerator card as "an architectural tweak" designed to boost GPU throughput.

Elsen said it is using its GPU-accelerated computing engine for applications ranging from pricing options and financial risk modeling to underwriting loans and testing actuarial assumptions. "Running more simulations increases both the quality of your results and your confidence in those results," noted Havell Rodrigues, Elsen's chief revenue officer.

For its part, Nvidia noted that the ability to run more complex financial models on greater data volumes—and do it in parallel—should translate into improved trading strategies and better risk management.

Elsen said it is making its scalable computing platform available as an API. The application-programming interface is based on a standard RESTful approach that leverages scalability and performance. The API is also billed as "data agnostic" and supports custom datasets.

Elsen is a new player in the emerging financial technology market focused on high-performance computing platforms for hedge funds, banks, insurers and other enterprises seeking more speed in running financial models and algorithms. Founded in 2013, Elsen seeks to differentiate itself by offering GPU-accelerated processing via either a secure cloud or on-premise.

EnterpriseAI