Memory performance

At all times and measurements, the system baseclock and multiplier will remain the same, meaning each and every difference you spot on performance is a direct results of changed memory frequency and timings.

Sandra Memory bandwidth

SiSoft Sandra has a fairly interesting memory benchmark suit which allows us to monitor memory bandwidth.

Today we'll be using an X58 chipset motherboard on which we plant a Core i7 965X processor. We'll leave this processor at reference clocks.

The only thing we'll be changing today are DIMM modules, frequencies and memory dividers. The Core i7 will run at all times at it's default clock, so that will not show any effect on performance. 

Okay so for the first round of tests, synthetic testing. Here we have the read performance of the memory kit at that 1600 MHz default and 1866 MHz (OC) frequency.

Then you need to compare to the very same setup, yet with JEDEC timed triple channel memory at 1066MHz - 3GB, 6GB at 1066MHz, we also threw in the results of a 6GB kit at 1600 MHz.

Most of these setups these days all run either 1066, 1333 or 1600 MHz depending on their baseline settings. Most Core i7 / X58 entries for example run triple channel memory configurations at 1066 MHz (default). Intel's Nehalem architecture and specifically here, the memory controller just hauls ass.

So all kits are triple channel, and as you'll observe, throughout the test session the Corsair memory obviously leads.

Memory Read test

Above AIDA at work with the memory read test. You'll notice that at 1600 MHz we reach almost 17 GB sec of bandwidth, which is stunning.

Memory Write test

Write performance then -- For comparisons sake, a Phenom II processor will perform roughly in-between 6500 and 7000 MB/sec -- with sharp timings you'd take it to roughly 7500 to 8000 MB/sec. In triple-channel mode Vengeance reaches an astonishing ~14400 MB/sec at 1600MHz.