PCMark05

We ran 3DMark06 because it is part of the Futuremark suite that is multi-threaded. This is also one of the reasons we chose PCMark05. Not only is PCMark05 multi-threaded, it includes two CPU tests that are exclusively multi-tasking performance tests. Running these will give us a good idea of the multi-tasking performance of these processors with respect to each other, as well as the overall performance of the processors. The CPU tests in PCMark05 are a very good way of getting a grasp on the overall computing power of your processor in the real world, as every test is based off of real world applications.


Here we see an even bigger difference than before. We’re betting that this performance increase -- and it’s a big performance increase at that -- is a result of those two exclusively multi-tasking tests we mentioned earlier. Those with a keen eye probably noticed some really weird numbers in the graph above as well. The QX6700 at stock scores the same as it does when overclocked to 2933MHz? Apparently; we ran this test dozens of times because we thought we were doing something wrong. We really don’t know how to explain this result. Also note the performance differences with the processors when they are running Prime95. You can see that the X6800 takes a much bigger hit than does the QX6700. And again, notice how the QX7600 performs significantly better than the X6800 at stock settings.

SuperPI (mod 1.5 XS)

The next test we chose to use was the venerable Super PI. Developed in 1994 by the Kanada Lab at the University of Tokyo, Super PI has been and remains to this day one of the simplest and clearest ways to test your computer’s sheer processing power. As has become common practice, the specific test we are reporting is the ‘1M’ test, or a test to see how fast your processor can calculate the value of pi out to one million decimal digits. This version of SuperPI has been modified by the gurus over at XtremeSystems.org to produce time figures that are accurate to the thousandths of a second. This allows us to pick up more subtle differences in computation times.


SuperPI is the first of our tests not to be multi-threaded. As a result, we wouldn’t expect the increased number of cores to have an effect, positive or negative, on the scores we achieved. However, you can see that the scores of the QX6700 at stock and at 2933MHz are significantly slower than those of the X6800 at stock. We don’t really have an explanation for this, other than that there could have been some sort of RAM influence going on. We will explain the whole RAM situation a bit later. It is easy to tell that when both processors were overclocked, the scores we achieved were pretty much the same, with no discernable advantage in either direction. You can also see that the addition of an instance of Prime95 slowed down the calculations pretty significantly as well, though much more so on the X6800 than the QX6700.