|G.Skill Sniper 1866 MHz DDR3 Memory Kit|
|Reviews - Featured Reviews: Memory|
|Written by Austin Downing|
|Tuesday, 14 June 2011|
Page 4 of 7
RAM Testing & Results
When testing RAM the preferred method is to remove as many bottlenecks as possible. This ensures that any changes in the benchmarks are dependent on the RAM that is being tested as much as possible. For this reason no games have been included because at the moment games are far more dependent on computers GPU's and CPU's, rather than their memory subsystems. Instead the tests used will be focused on more memory intensive tasks such as fluid dynamics simulations, compression, and real-time rendering. These benchmarks will be run alongside traditional benchmarking tests that will scale more effectively with speed but will be an unrealistic indication of real world performance.
In order to test each RAM kit in the most effective fashion a multistep process was required. First, in order to verify the advertised speed each set of RAM they will be initially run the Memtest86+ for one run. Once no errors are found, Windows 7 Professional 64-bit with all of the need tools will be loaded. In order to make sure that fluctuations in other subsystem do not influence the scores, each test will be run a total of three times. At the end of the benchmark the lowest score will be dropped and the final two scores will be averaged. The applications being used for benchmarking will be as follows.
Because each IC produced is slightly different and will have different overclocking characteristics any overclocking results obtained may vary from user's actual experience. In the end this means that spending the time to overclock our system memory will be unbeneficial to our readers and therefore will be omitted from the results except for in special cases, such as when a product is advertised as having a great amount of headroom.
We have recently introduced the Euler3D Computational Fluid Dynamics benchmark into our memory test suite and so I would like to give a small explanation as to what is being used. Our newest test simulates the fluid dynamics of a wing and therefore is very memory and CPU intensive. Each test is run and the final result is given in Hertz(Hz). For our purposes the final Hertz score is what will be used for comparing each set of memory to its competitors. It has been found that this benchmark is particularly sensitive to both memory speed and timing and therefore makes a perfect benchmark to add to Benchmark Reviews memory suite.