October 3, 2013

‘Ivy Bridge’ Up to 1.5X Faster than ‘Sandy Bridge’

Tiffany Trader

The solution architects at Xcelerit have performed another series of benchmark tests, this time comparing the newly released Intel Ivy Bridge processors with the previous-generation Sandy Bridge parts. The 12-core Ivy Bridge (Xeon E5-2697 v2) chips will be tested against the 8-core Sandy Bridge (Xeon E5-2670) using a financial application – a Monte-Carlo LIBOR Swaption Portfolio pricer. The implementation employs the Xcelerit toolkit, which allows compute-intensive applications to benefit from hardware accelerators (GPUs and multicore) with only minor modifications to the source code.

As the Xcelerit blog explains, the biggest difference between the Sandy Bridge and Ivy Bridge line is the processing die shrink from 32nm to 22nm. The reduction in the production process allows more transistors to fit on a chip and enables higher clock rates. In other respects, the designs are very similar, although it’s worth noting that Ivy Bridge’s 22nm 3D technology lowers power consumption by as much as 50 percent.

The test application comes from the world of quantitative finance. It’s a Monte-Carlo simulation that is used to price a portfolio of LIBOR swaptions (financial swap contracts). The blog explains that “thousands of possible future development paths for the LIBOR interest rate are simulated using normally-distributed random numbers.” Each development path represents one Monte-Carlo path.

The setup for the experiment is simple. The Xcelerit team runs the same Monte-Carlo application on both the Sandy Bridge and the Ivy Bridge-based servers and compares the outcomes. The test systems, each with two CPUs, are configured with the following specs:

CPU: 2 sockets, Sandy-Bridge (Xeon E5-2670) or Ivy-Bridge (Xeon E5-2697 v2)

  • HT: Hyper-threading enabled
  • OS: RedHat Enterprise Linux 6.4 (64bit)
  • RAM: 64GB
  • Compiler: GCC 4.4
  • Xcelerit SDK: version 2.2.2

The total run time takes into account the execution time of the full application, inclusive of data transfers, random number generation and reduction (part of the algorithm).

Paths Speedup Ivy-Bridge vs. Sandy-Bridge
64K 1.15x
128K 1.25x
256K 1.34x
512K 1.4x
1,024K 1.48x

The results show that for all number of paths taken, Ivy Bridge outperforms its predecessor. In a linear fashion, the higher the number of paths, the greater the speedup, all the way up to 1.5 speedup for the highest number of paths. The Xcelerit team explains that this reflects the increase in cores from 8 to 12 per chip, stating: “The benefits of the new Ivy-Bridge for financial Monte-Carlo applications can be clearly seen here.”

Related Items

Phi and Kepler Run Monte Carlo Race

Intel Intros Xeon Processor E5-2600 v2 Product Family

GPUs Show Big Potential to Speed Pricing Routines at Banks

Intel Xeon Phi Versus ‘Sandy Bridge’