In a dual-socket configuration with 24-cores total and running the same code as was previously run on Intel’s octal-core “Sandy Bridge” Xeon E5-2690 processors, the latest 12-core “Ivy Bridge” Xeon E5-2697 v2 processors are 50 percent faster than the previous generation, the Securities Technology Analysis Center LLC (STAC) found after running its financial benchmarks on the new chips.
Called the STAC-A2 benchmark, the software stack was coded by Intel using its Composer XE (revision B), its Math Kernel Library (MLK) and its C++ compiler. Parallelism was handled using Intel’s OpenMP 5.0, all running on a white-box chassis with 128 GBs of random access memory (RAM) under the Red Hat Enterprise Linux 6.3 operating system.
The STAC-A2 benchmark is a so-called “Greeks” benchmark for risk management.
“Risk management is a big deal in banking these days,” said Peter Lankford, founder and director of STAC. “Banks need to run simulations in order to quantify their risk. STAC-A2 is based on market risk, which is what’s likely to happen to the value a portfolio if the market moves one direction or another. Running these simulations takes a lot of computing power, consequently there is a lot of attention on how much you can do in what unit of time and with what amount of energy. STAC-A2 is a suite of benchmarks based on options Greeks, which tell you the sensitivity of the value of an option based on other kinds of changes, such as changes in interest rates.”
These risk sensitivities–sometimes called hedge parameters–are called Greeks because they are often represented by Greek letters. For instance, for the most common types of options–called vanilla options–the Delta Greek is a measure of the rate of change of an option value with respect to changes in the underlying asset’s price. The Delta Greek is measured by taking the first derivative of the option value with respect to the underlying instruments price–a computationally intensive operation. Other examples include Vega–a measure of sensitivity to volatility, Theta–a measure of sensitivity to the passage of time and Rho–a measure of sensitivity to the interest rate. There are also whole sets of second- and third-order Greeks that are even more computationally intensive since they calculated with second- and third-order derivatives, respectively.
“STAC-A2 is the ultimate end-to-end benchmark, where you are required to calculate all these Greeks,” said Lankford. “But it also breaks things down and tests the components–testing each Greek individually–as well as the components operations such as square roots and generating random numbers, so that the user can understand where the performance is within a Greeks operation, and also how a particular system that we test may perform in other kinds of algorithms besides these Greeks.”
The bigger banks have thousands of computers configured in grids running these types of calculations every day, which was traditionally done overnight in batch mode, but is increasingly shifting to intra-day or, if possible, in real-time–which Lankford defined as less than one second processing time.
“One of the important things about STAC-A2 is that it gives you a sense of scaling–in addition to the raw performance benchmarks,” said Lankford. “And the other thing is that it gives a lot of focus on the quality of the algorithms submitted by the vendors.”
To view all the details of STAC-A2 and the results obtained for Intel’s new Xeon E5-2697 v2 processor visit the STAC website. Also available there are the results for STAC-A2 running on Intel’s Xeon Phi coprocessor.