SunGard Adaptiv Credit Risk
The SunGard Adaptiv Credit Risk (ACR) benchmark is actually a version of an application from SunGard (www.sungard.com) that was outfitted with a GUI front-end by Intel. ACR uses a proprietary Monte Carlo simulation engine to analyze the risk and return on a hypothetical portfolio of assets. The benchmark portfolio consists of 1428 deals, mostly interest rate swaps and foreign exchange forwards; the portfolio is examined in conjunction with 29 price factors and 19 risk factors. The application decomposes the work across the portfolio, depending on the resources available, so there is a fair amount of parallelism, although some parts are serial or require replication.
While this may sound esoteric, applications like this are extremely common in the financial services industry. Credit risk analyzers are used to issue credit cards, loans, mortgages, bonds, etc. by banks. They are also used by third parties for trading the aforementioned instruments. These applications are considered mission critical by many of the world largest financial companies, including Fannie Mae and Freddie Mac, where they are used to analyze the risk of mortgages.
Unlike the rendering benchmarks, the SunGard benchmark has much finer threading control. The user may set the number of threads at initialization, so that both performance and scaling can be observed. When a Hyper-Threading aware OS is used, preference will be giving towards physical processors. This means that the speed up from Hyper-Threading can be directly measured as the difference between two and four threads for the Nocona system and four and eight threads for the Bensley system.
Figure 4 – SunGard Adaptiv Credit Risk Performance
The standard deviations for the Nocona run times were 0.85, 7.33 and 0.66 seconds for 1-4 threads. For Bensley, the deviations were 2.99, 5.16, 2.14 and 2.00 for 1-8 threads respectively. The Nocona system scales by a factor of 2.44, and the Bensley system by a factor of 4.81. The Bensley system at eight threads is 88% faster than the Nocona.