For the purpose of testing I will be using the G Skill Sniper kit and all subsequent Sandy Bridge memory on the ASUS P8P67 motherboard, using the Intel 2500K. Our database is now growing for Sandy Bridge memory reviews, so over time there will be more memory to compare against. But for now, we will be testing versus OCZ Reaper, G.Skill RipjawsX and recently reviewed Kingston HyperX Special Edition's.



Overclocking

With BCLK being fixed on the Sandy Bridge platform this has now altered the way we can overclock. Overclocking power with Sandy Bridge is concentrated on the CPU rather than memory. We do not have the flexibility we once had by changing the BCLK in order to move the frequencies in small increments, instead we have only set margins to move in because of the divider that is set. So available we have 1066MHz / 1600MHz / 1866MHz / 2133MHz / 2400MHz. In essence we have little room to experiment with, but lets see what we can do.

So to begin with I overclocked the kit. With the intention of pushing the memory frequency up to the first milestone - 1866MHz. It's always a good idea to loosen the CAS timings when overclocking so that nothing affects moving up to the specified frequency. So I slackened the timings to 10-10-10-28. Immediately I ran into problems, so I increased the voltage from 1.25v to 1.6v incase this was hindering the frequency increase. Sadly, although the Sniper kit did boot into Windows and some of the benchmarks did run. It wasn't stable and 3DMark 11 and SANDRA revealed instability. All of our previously tested Sandy Bridge kits managed to move up from 1600MHz to 1866MHz without any problem of stability.

The next objective was to modify the CAS timings at the stock memory frequency of 1600MHz. The idea here is to get the timings as tight as possible and the lowest ratings that I could settle on were 8-9-8-25 @ 1.5v. So I was able to move the kit down to CL8, with a slight modification to the voltage.

With the performance modifications being established I was then able to begin benchmarking. So the Sniper kit will be tested in CL8 and at CL9 (stock) ratings, sadly due to the stumbling block with overclocking we won't be able to test the kit with an uprated frequency setting. The results are on the proceeding pages. To summarise here are the settings we will be testing on the G Skill Sniper's: