Performance in Nominal Mode and During Overclocking

Since EVGA X58 SLI Classified mainboard is the only one besides Gigabyte solutions that could overclock our test processor to its maximum frequencies, it would be absolutely natural to pick one of the Gigabyte boards for performance comparison. However, unfortunately, we won’t be able to offer you a comparison like that today. Gigabyte GA-EX58-Extreme mainboard that we had at our disposal passed all the tests in nominal mode, but for some reason refused to boot once we increased the memory or base frequencies. I took the entire testbed apart and then put it back together three times, but everything stayed the same: even when we raised the base frequency from 133MHz to only 134MHz, the board announced “over-overclocking” and reset all parameters to their nominal values after a few failed reboots.

In fact, the board didn’t die as an overclocker solution. We have recently confirmed that it still does overclock processors and memory just fine on a different testbed. Therefore, we are going to go back to this matter a little later and try to find out what was the reason behind this strange problem with our Gigabyte GA-EX58-Extreme mainboard. And in the meanwhile let’s compare the performance of EVGA X58 SLI Classified against Asus P6T mainboard that we have reviewed a while back. The first chart will represent the results obtained in nominal mode, when nothing is overclocked and all parameters are set by the mainboards themselves:

Well, we could probably consider a 1-3% difference to be within the measuring error margin, even though we repeated each test multiple times, discarded the results that fell out of line and calculated the average for the charts and further analysis. However, the SuperPi results are usually very stable and 8 seconds for calculation of 167 million digits is the worst result we have ever got. No other system based on Intel Core i7-920 processor and any of the mainboards on Intel X58 Express chipset has ever been so slow here. Besides, more that 7% difference in FarCry 2 and almost 12% in Image Editing benchmark fall way beyond the acceptable error margins.

And what do we have during overclocking? We overclocked our processor to 3.8GHz on Asus P6T mainboard, and that was the best we could get. As for EVGA X58 SLI Classified, we decided to take overclocking to 3.9GHz frequency. Of course, the board could overclock even higher, to 3.95GHz. However, 50MHz is not that much of a difference, while we lose in the memory frequency and don’t have Intel processor power-saving technologies working properly. These are the reasons why I consider overclocking to 3.9GHz to be a more optimal choice here. Let’s check out the results now:

We can explain a slight advantage of EVGA X58 SLI Classified mainboard in Cinebench 10, Fritz Chess Benchmark, SuperPi 8M and Crysis Warhead by 100MHz higher CPU clock frequency. However, there is barely any difference in 3DMark Vantage, the board starts falling a little behind the competitor in FarCry 2, and in Image Editing benchmark the gap increases enormously again. So, what’s the matter? If you think about it, you may come up with at least three reasons.

Here is the first one. In fact the “standard” overclocking of our particular Intel Core i7-920 processor sample by raising its base frequency to 181MHz, which most mainboards do even without increasing their Vcore, is very good from two standpoints. First of all, we all remember about Intel processor power-saving technologies that remain up and running. But besides that, this particular mode appeared great for Kingston HyperX KHX14900D3T1K3/3GX DDR3 SDRAM. When we first talked about this memory, we learned that by setting CAS Latency 8 you could get these modules to work at maximum 1812MHz frequency. Overclocking your system to 181MHz base frequency allows setting the memory frequency very close to this maximum for CL8. At 186MHz base, the memory frequency is a little – 50MHz – higher, however, in this case we have to increase the timings and set CL9. We do get a performance boost in some applications due to higher processor and memory frequencies, but lose in a few others because of higher memory timings.

The second possible explanation is the following. We compared performance of two mainboards, but in fact, there was no direct comparison. The results for Asus P6T were taken from our earlier test session when we compared it against Foxconn BloodRAGE. As for EVGA X58 SLI Classified, we took the performance readings not on the freshly installed OS, but on the OS used for our DFI LanParty JR X58-T3H6 tests. There should formally be no noticeable difference, because all settings are the same, just as the list of test applications. But when it comes to principal comparisons like this, every trifle counts.

And finally, there is one more hypothesis to be considered. Maybe EVGA X58 SLI Classified mainboard is simply slower than Asus P6T?