NVIDIA Grace Hopper Superchip Boosts Murex MX.3 Performance, Enhances Energy Efficiency
NVIDIA’s new Grace Hopper Superchip is transforming the capabilities of Murex’s MX.3 platform, significantly enhancing performance metrics while reducing energy consumption, according to the NVIDIA Blog.
Managing Risk With MX.3 Driven by Grace Hopper
Following the 2008 financial crisis, regulatory changes have pushed financial institutions to adopt more robust risk management frameworks. Pierre Spatz, head of quantitative research at Murex, anticipated a growing demand for more efficient computing solutions. In response, Murex has adopted NVIDIA’s cutting-edge technologies, including CUDA and GPU-accelerated computing.
Recently, Murex began testing NVIDIA’s Grace Hopper Superchip, aiming to enhance the performance and energy efficiency of its MX.3 platform. MX.3 is used by over 60,000 daily users across 65 countries for trading, risk management, and operations in various asset classes.
The Grace Hopper Superchip is specifically being tested for its ability to handle X-valuation adjustments (XVA) and other critical calculations. Murex has reported that the superchip delivers faster calculations and significant power savings, which are crucial for institutions needing to run complex risk models on large datasets.
“On counterparty credit risk workloads such as CVA, Grace Hopper is the perfect fit, leveraging a heterogeneous architecture with a unique mix of CPU and GPU computations,” stated Spatz. “On risk calculations, Grace is not only the fastest processor, but also far more power-efficient, making green IT a reality in the trading world.”
Early tests show that the Grace Hopper Superchip can reduce energy consumption by 4x and improve performance by 7x compared to traditional CPU-based systems when running XVA workloads in MX.3.
Pricing FX Barrier Options in MX.3 With Grace Hopper
In addition to risk calculations, Murex has leveraged the Grace Hopper Superchip for pricing foreign exchange (FX) barrier options using its advanced stochastic local volatility model. A barrier option is a type of derivative with a payoff dependent on whether the price of the underlying asset crosses a specified threshold during the contract period.
The pricing evaluation involves solving a two-dimensional partial differential equation more efficiently on the Arm-based NVIDIA Grace CPU in the GH200. Testing shows that the Grace Hopper Superchip processes these calculations 2.3x faster than Intel Xeon Gold 6148 systems. Furthermore, the chip is 5x more power-efficient on a per-server basis.
NVIDIA’s commitment to energy-efficient, high-performance computing is evident in these advancements, providing substantial benefits for quantitative analytics in the financial sector. Murex’s implementation of the Grace Hopper Superchip is a testament to the accelerated computing platform’s capabilities in improving cost-efficiency and computational speed.