Overclocking

I had a lot of fun overclocking these sticks on the Intel X79 Express platform. I reached a maximum speed of 2427 MHz, higher than any other kit to date. I confirmed that these sticks do use the same memory ICs as the 16 GB 2133 MHz CL9 G.Skill kit I reviewed a while ago, and so I changed up the secondary timings to match what I used with the G.Skill sticks, and that provided me with some extra performance as those timings are more optimized for this platform in comparison to the stock timings that were really optimized for the P55 Express platform.

All of our testing, even all the way up past 2400 MHz, used just 1.05 V set to VCCSA, even though there are many guides out there on other websites suggesting setting up to 1.25 V for VCCSA, and sometimes up to 1.2 V for VTT. There are some discussions on sites like HWBOT that are currently investigating some degredation that some users have noted over short periods of time using such voltage settings, and this degredation is something I have personally noticed with my ES i7 3960X CPU. Although at this time I cannot exactly pinpoint what voltage caused the damage, our testing CPU has lost some of the maximum CPU overclock it used to have, but it doesn't seem to have any impact on memory scaling. This means that I was forced to change cooling solutions, and I now employ a Corsair H100 all-in-one watercooler, as our CPU now requires more voltage for the same clocks than it did just a couple of months ago. Now, because it's only CPU speed that is affected, and not BCLK or memory speeds, I feel that the VTT voltage may be the one that is at fault for the degredation, and as such, I suggest that if you do invest into the X79 Express platform, and are overclocking, that you do not exceed more than 1.15 V on the VTT, as this seemingly does provide more than enough for most CPUs to reach a high BCLK setting, including my degraded chip.

I know this may not seem to relate exactly to the testing at hand, or to these Patriot sticks, but because the cooling solution I use during testing has changed, I do need to make that obvious, in order to remain fully transparent. I can in no way confirm or deny the possibility that certain overclocking or settings may cause damage to your own parts, nor can I say 100% for sure what damaged my own parts, but there does seem to be a trend emerging as noted by other reviewers and those that play in the extreme-clocking scene, and this may be a subject that needs to be addressed directly in the future. I just wanted to highlight that although these sticks are not intended for this platform, and do require a fairly hefty BLCK overclock to reach "default" speeds, the required settings and voltages to do so do not land within what I like the call "The Danger Zone" of overclocking on this platform. As always, you may be taking certain risks when overclocking, and while those risks may not be immediately known when a platform launches, with the Intel X79 Express platform being so new on the market, I do advise some caution when pushing things to the limit.

Of course, the changed cooling might affect memory clocking a little, so I did go back and retest the G.Skill sticks, but they refused to go any further than they had before. That said, the Patriot PX538G2000ELK kit went further than any other so far, but because of a difference in timings used compared to the G.Skill kit, those top clocks may not equal top performance, and I am not 100% sure if it's the timings, or the lack of four sticks versus two that led to the results below. Hopefully as I add new numbers and kits to the testing a more clear picture of what is exactly happening will emerge.

Overclocked Performance Summary

SuperPi finished a total of 11.35 seconds faster with the 427 MHz overclock, pretty respectable.


wPrime got a decent boost as well, but it only amounted to 0.36 seconds overall.


AIDA Read Performance got close to a 1200 MB/s boost, and left the Patriot PX538G2000ELK kit just 746 MB/s behind the G.Skill kit.


Latency improved by 4.5 ns, and this time the G.Skill kit lost out to the PX538G2000ELK kit.


WinRAR also got a good boost, just 8 kB/s behind the G.Skill kit, well within the margin of error.


Shogun 2 finally showed about a half and FPS worth of gains versus the XMP profile, not a lot, but still measurable.