For today's testing we will be using the Gigabyte EX-58 UD5, a mid-range Core i7 motherboard from Gigabyte that will allow us to push the memory on test to its absolute limit. Here's a breakdown of the rest of the components:
Intel Core i7 920 'Nehalem' @ 2.66Ghz
DFI Lanparty T3eH6 DK
Kingston Value CL9 1333MHz 9-9-9-24 6x2GB kit
Kingston Value CL9 1333MHz 9-9-9-24 6x2GB kit (Overclocked to 8-8-8-24 @ 1600MHz)
Gigabyte Odin 1200w
Windows Vista Ultimate 64bit SP2 + Updates
For testing the memory we used a number of synthetic benchmarks and games:
- Lavalys Everest 4.10
- SuperPI mod_1.5
- Sisoft Sandra 2009
- Passmark (Memory)
- 3DMark 05
- 3DMark 06
- 3DMark Vantange
- Call of Duty 4
- Far Cry 2
Starting from scratch we disabled on the settings that may affect the overclocked settings such as Intel Speed Step as well as disabling the C-State settings which may also affect some of the results in the benchmark testing phase of the review. Here's how the sticks look at stock speed:
I did not expect an significant overclock from the Kingston kit being that they are aimed at the OEM system builder and not designed nor marketed towards the overclocking sector but I did manage to squeeze a little extra out of the kit:
First I tried to overclock the kit with everything set to there stock settings and just decreasing the memory divider. This resulted in a non boot scenario so I had a play with the voltages and settled for 1.6v Vdimm, 1.36Vtt and 1.35Vcore on the CPU. This not only allowed a break through to 1600MHz but also allowed the timings to be lowered to 8-8-8-24 from 9-9-9-27. I also managed to run the kit at 1T rather than the stock 2T. The Elpida chips certainly seemed to respond well to this small bout of voltage increases. Sadly, any further increases seemed to decrease stability so this was pretty much the maximum speed I could get out of the kit without enduring the dreaded BSOD's so often associated with over exuberant memory overclocks.
Returning the kit back to stock, I also tried to lower the timings but again I ran into stability issues unless I increased the voltages managing the same latencies as I did with the memory clocked to 1600MHz. Regardless, this is a great result for what is in effect budget memory, especially when you consider that the kit on test was run in a 12GB configuration.
Let's see how the memory performs in our suite of benchmarks...