ASUS Rampage II Extreme X58 Motherboard Review

Overclocking Results

The above says it all; overclocking is why the Rampage II Extreme was created, and it is why people will consider spending top dollar on this model instead of lesser boards. Now let's see if it is truly worth the premium...

...but first, keep in mind that overclocking Core i7 processors is quite different than with the previous Core 2 architecture. There are now five clock speeds (CPU/BCLK/MEM/UCLK/QPI) and four multipliers (CPU/MEM/UCLK/QPI) to tweak, as well as eight different voltages. Put simply, there are lot of new variables and potential limitations that an overclocker must now take into consideration. At the moment, no one is an expert when it comes to Core i7 overclocking, and we are all learning new tips, tricks, and tweaks on a nearly daily basis.

Before beginning, you should enter the CPU Configuration menu in the BIOS and disable the CPU TM Function. This will circumvent the "Overspeed Protection" mechanism that limits current draw and CPU power consumption to 100 A and 130 watts, respectively. You are unlikely to ever encounter these limits, but we might as well remove any potential obstacles right off the bat. Secondly, after selecting your BCLK and memory frequency, always set the UCLK and QPI speeds as close to default as possible, in order to prevent them from becoming overclocking bottlenecks. This is particularly important because at the moment it appears that the QPI frequency is the single greatest limitation to Core i7 overclocking (perhaps second only to temperatures), since anything above the 7600-7800MT/s (3800Mhz-3900Mhz) range is nearly impossible to accomplish with any stability. Thankfully, Intel are well aware of this fact, and they are working on providing lower QPI multipliers as we speak.

On similar note, as stated in the BIOS the UCLK must be at least twice the DRAM frequency. Some have argued that the UCLK should be lower than the QPI frequency, but we have had no stability issues when running the UCLK equal to or even above the QPI rate.

Some people will argue that HyperThreading should be disabled since it uses more power and causes higher temperatures, but we will leave it enabled in all our tests since we believe it is a central element of the Core i7's appeal.

Please note that CPU-Z misreports the memory timings, therefore rely on Everest instead. Also, don't pay attention to the HyperPI times, when running 8 processes the gap between the fastest and slowest run is huge (3-6 minutes), so it is simply used for stability testing purposes.


Highest Stable BCLK Overclock


As you can see, our high expectations were not misplaced, the Rampage II Extreme allowed us to increase the base clock (BCLK) from 133Mhz to 211Mhz, which is almost a 58% increase. Quite impressive for a first-generation motherboard. At the moment, it is generally accepted that the stable BCLK range is between 180Mhz and 210Mhz, so the fact that we were able to achieve the high-end mark reflects very positively on the RIIE.

In order to achieve this result, we had to increase the QPI/DRAM voltage from 1.20V up to 1.50V, which is tentatively the highest setting that we would recommend for extended benchmarking use. In order to consistently load Windows, we did have to increase IOH and ICH voltages a few increments.

Although CPU overclocking is not the focus of this section, we were able to overclock our Core i7-965 XE to 4.0Ghz, a respectable 25% increase. This overclock required approximately 1.40V, which is once again the highest setting that we would recommend for 24/7 use, if only because it was extremely difficult to keep CPU temperatures below 80C at full load. Having said that, our chip is an engineering sample, and we have seen some retail Core i7 processors with the C0/C1 stepping that not only overclock better than our E.S sample but with much less voltage as well.

By the way, just to demonstrate how truly fast this CPU Clock/BLCK combination is, click here to see a proper single SuperPI Mod 1.5 run.


Highest Stable Memory Overclock


As you can see, our 6GB G.Skill DDR3-1600 memory kit was able to achieve 1758Mhz 9-9-9-24 (remember CPU-Z is misreporting the memory timings). While this result will not impress those used to the DDR3-2000+ speeds achieved on the Core 2 platform, it is nevertheless a decent result for a first-generation low-voltage triple channel memory kit. Although we did achieve this overclock with the "recommended maximum" 1.65vDimm, this was simply because the kit did not respond to higher voltage at these timings.

With respect to the much publicized voltage limit, we can confidently state that all the hype was for nothing. Yes, very early engineering samples were susceptible to damage when high Vdimm was used, but the retail chips are much more robust. What matters the most with respect to this issue is the voltage differential between the QPI/DRAM voltage and the DRAM voltage, the general consensus is that it should not be larger than 0.5V. The QPI/DRAM voltage feeds the integrated memory controller, so it is obviously one of the key elements in memory overclocking on this new platform. For example, in order to run our memory kit at its default DDR3-1600, the QPI/DRAM voltage had to be increased from 1.20V to 1.35V (the RIIE will do it automatically as well), tweaking the DRAM voltage made no difference in stabilizing this memory frequency.


Overall Stable System Overclock


For our overall stable system overclock, we selected a 4.0Ghz core clock, 200Mhz BCLK, and DDR3-1600 memory speeds.This effectively mimics the clock speeds of a Core i7-920 overclocked to 4.0Ghz, which is what we would run on a 24/7 basis, and it will serve as the overclocked configuration in our subsequent benchmarking section. Before anyone asks, we did not bother overclocking our memory kit because triple-channel DDR3-1600 already provides a huge amount of memory bandwidth, much more than our overclocked Core i7 needs, and higher memory speeds simply did not result in worthwhile performance improvements.