System Builder Marathon: Low-Cost System

Test System & Benchmarks

Although we've just had a detailed analysis of the components we chose, let's scrutinize them one last time before moving on:

We experienced a few peculiarities when putting the low-cost system together. Nothing earth shattering, but we'd like to share them with you.

First, Windows Vista installed, but the system entered an infinite loop on the first boot, with Vista displaying the crawling loading bar at the bottom of the screen. This problem disappeared when we set the SATA boot drive to IDE mode.

Second, we weren't interested in lowered performance, so we turned off the 'Phenom TLB Erratum patch' in the BIOS. For those of you who don't know, AMD's first batch of Phenom processors have a bug that, in extremely rare conditions, can manifest itself in a crash. This bug is called the 'Phenom TLB Erratum', and has been fixed with a patch in the BIOS of most Phenom motherboards. Unfortunately, since the patch shuts off the Phenom's important L2 cache it results in a notable performance drop.

On some boards - like our Gigabyte MA78GM - this patch can easily be turned off in the BIOS, allowing the Phenom to work at its full potential, which is what we did. Those of you who find this disturbing should know that the Phenom's TLB erratum is so rare that we've never seen it manifest itself in testing, and the system never crashed once while we subjected it to our punishing benchmark suite.

The closest thing we found to a problem that with the system was that when AMD's Cool & Quiet technology was enabled in the BIOS, it didn't seem to work properly. Cool & Quiet is supposed to lower CPU clock speeds when the system is idle to reduce power consumption and heat output, which is great, but after a great deal of testing we realized that Cool & Quiet was not ramping CPU speed properly under load. In some cases we ran encoding benchmarks and noticed that the CPU was running at half the advertised 2.2 GHz speed under load! The fix for this was to simply shut off the 'Cool & Quiet' feature in the BIOS, which resulted in over 30% faster performance in most of the benchmarks we had already taken.

Below are the test settings we've used for our myriad benchmarks and tests. We have completely revamped our test suite this time, so unfortunately we won't be comparing results with the previous builds until the next system builder marathon. However, all of the systems in this marathon will be using the same benches and settings so we can compare the results between PCs later in the week:

Now that we know what settings we've used, let's have a look at the results. These will become even more interesting when we have results from mid- and high-end systems later this week.