Performance

Now it is time to compare the performance of Asus and Gigabyte mainboards. We will start our tests with the nominal operational mode when the mainboards adjust all their settings automatically. We have included the SuperPI test into the list of benchmarks on purpose: it will show the advantage of a more “correct” Turbo Boost implementation on Asus P6T (this technology increases the clock multiplier to 22 in case of single-thread workload). Let’s check out the results:

The mainboards performance is overall similar. Despite our expectations, Asus mainboard is behind Gigabyte in SuperPI test. However, it performed unexpectedly well in synthetic Everest memory tests. What’s the trick? Does Asus know some secrets about memory work on an Intel X58 Express based platform, or it all comes from increasing the multiplier to 22 vs. 21 on Gigabyte boards? We can easily find out the truth. Let’s limit the maximum multiplier setting on Asus board to 21 as described above and repeat the tests:

The results have leveled out, which means that during the first round of tests Asus benefitted from a higher x22 multiplier. Unfortunately, more “correct” Turbo Boost implementation on Asus P6T is not correct enough to guarantee performance advantage in real applications. If the theoretical advantage doesn’t find any practical use and can only be noticed in synthetic tests, then it is not really an advantage anymore. Unfortunately, Turbo Boost is implemented incorrectly on all mainboards we have tested so far including solutions from Gigabyte, ASRock and Asus.

Now let’s compare the mainboards performance during processor overclocking. We have often seen mainboard performance drop during overclocking, even though it performed as fast as the competitors in the nominal mode. To ensure that it would be a fair comparison, we overclocked our processor to its maximum on Asus P6T. As for Gigabyte GA-EX58-Extreme, we couldn’t let it go full throttle on us, even though it can do more. We overclocked our test CPU to 3.8GHz, the memory was working at 1810MHz with 8-8-8-22-1T timings.

Luckily, there were no surprises this time. The boards performed equally fast. Now let’s see what price we have to pay for these results.