Pierre Spatz, head of quantitative research at Murex in Paris, foresaw a shift towards cost-cutting in banking post the 2008 financial crisis and tighter regulations.
To achieve optimal efficiency, Spatz embraced NVIDIA’s CUDA and GPU-accelerated computing at Murex.
Keeping up with advancements, the quants team at Murex has initiated trials of the NVIDIA Grace Hopper Superchip to enhance services for pricing and risk management of derivative contracts.
Over 60,000 users across 65 nations rely on Murex MX.3, which supports various financial institutions like banks, asset managers, and pension funds with trading, risk, and operational needs.
Enhancing Risk Management with MX.3 Driven by Grace Hopper
Financial organizations require robust computing infrastructure to analyze massive datasets for risk and pricing, and to enable real-time decision-making.
MX.3 covers credit and market risk, BASEL capital standards, trading book review, and x-valuation adjustment (XVA) among others.
Grace Hopper is being tested at Murex for XVA and market risk calibration, pricing, sensitivity analysis, and profit evaluation across different asset types.
The Grace Hopper chip not only accelerates calculations but also contributes to power efficiency on the Murex platform.
Regarding counterparty credit risk tasks like CVA, Spatz mentioned, “Grace Hopper is well-suited, leveraging a hybrid architecture combining CPU and GPU computations. It not only outperforms other processors in risk calculations but also significantly saves power, promoting eco-friendly practices in the finance industry.”
In XVA workloads on MX.3, Murex R&D labs have observed a 4x decrease in energy usage and a 7x performance boost compared to CPU-based systems.
Valuing FX Barrier Options in MX.3 With Grace Hopper
Murex employs the latest stochastic local volatility model to price FX barrier options, witnessing notable performance enhancements on Grace Hopper. Barrier options depend on specific asset price thresholds during the contract period.
By computing a 2D partial differential equation, costing is more effective on the Arm-based NVIDIA Grace CPU in GH200, resulting in 2.3x faster pricing than Intel Xeon Gold 6148.
NVIDIA Grace CPU also delivers notable energy efficiencies for FX barrier calculations compared to watts-per-server basis, exhibiting a 5x improvement.
Murex acknowledges that NVIDIA’s next-gen accelerated computing platform is enhancing energy efficiency and lowering costs in high-performance quantitative analytics for capital markets.
Explore NVIDIA AI solutions for finance industry.
