Test results

All the motherboards are actually based on the same chipset, which makes testing them much easier: there is no need to test various operating modes. The only exception could have been made for integrated video in i945G. But we already reviewed the performance drop caused by GMA900/950 as well as absolute performance results of these video accelerators, so we can do without repetitions. We used our standard pair of DDR2-533 modules from Corsair, which SPD contains 3-3-3(-8) timings for 533 MHz. We reduced timings to minimum in each case. All the contenders underwent this test successfully, working well with 3-2-2(-4) timings, which is the theoretical minimum. Well, is there no plot in this roundup, diagrams demonstrating five equal bars? Let's check it out.

While both Foxconn motherboards demonstrated nearly the same memory read/write speed and Gigabyte 8I945P-G writes to memory slower than reads from it, both models from ECS demonstrate evidently low performance.

Let's have a look at the maximum memory exchange rate (using prefetch and non-temporal store methods). Gigabyte motherboard slightly outperforms its competitors — that's all right, considering that its FSB frequency (and thus CPU clock) is higher by half percent. In this test, all other things being equal (processor + memory), all the models must have demonstrated similar reading results and absolutely the same (to within FSB frequency) writing results. But we can see that both motherboards from ECS are noticeably outperformed. Considering their startup problems with Pentium 4 EE (1066 MHz bus), problems in our tests are not quite like a bolt from the blue, but we still didn't manage to find out the reasons.

The situation with memory access latency is also quite expected: ECS PF5 Extreme and ECS 945P-A (2.0) are much slower, the other three models are on a par. Foxconn models are slightly outperformed, leaving hope for Gigabyte 8I945P-G to make up for its reduced write rate with the other parameters and fight for leadership in real tests.

Indeed, Gigabyte motherboard turns out several seconds faster in archiving than the pair of Foxconn models that demonstrate the same results. What concerns the "unfortunates" from ECS, they are slower by up to 10%.

Interestingly, even the video encode speed (measured according to our method) demonstrated by ECS motherboards is a tad lower, though these results usually don't depend on a memory operation speed.

Both models are also a couple percents slower in image processing in Photoshop, which is also not very critical to memory performance.

So the situation is crystal clear, but we shall also publish results of our game tests: these applications illustrate the difference in synthetic mode (640×480 and 800×600 with low graphics quality on a powerful modern video accelerator), which as a rule disappears in real conditions.

Testing the integrated audio quality

The integrated audio quality of each model was tested in 16bit, 44 kHz using the RightMark Audio Analyzer 5.4 test application and the Terratec DMX 6fire sound card: