Page 6

Tests
As stated we are a hardware site with an audience that is primarily gamers, therefore we will review CPU's in a somewhat different way then other sites tend to do, mostly with gamers benchmarks in various resolutions. Not because we like them so much, no because I know you want to see that. We could test how fast your spreadsheet's macro function is updated or if that mp3 was encoded 2 seconds faster but I know for a fact you don't care very much about that. We'll go in-depth the gamers way. We will also test games under normal settings and conditions.

Let me explain, the best way to produce benchmarks scores in a gaming environment for a CPU is to lower the screen resolution and measure at 16-bit. Why you ask? Well, the graphics card bottleneck is a key issue here. At a certain point your graphics card becomes the more important and dominating factor to influence overall performance. I believe that the CPU and graphics card are a symbiosis though. There's only one important rule I uphold strongly and that is to use one of the fastest graphics cards available to date and use that to produce results. Do you care how CPU's rate at 640x480 @ 16 bit ? No, we don't either, therefore we are testing hardware the way you play it at home as well.

SiSoft Sandra was used to measure CPU and Memory Bandwidth, while MadOnion's PC Mark 2004 was also used to measure memory bandwidth and CPU performance. Furthermore, Splinter Cell, Half-life 2 and a several OpenGL and Direct3D games like Doom III were used to measure real-world multimedia performance.

Now we begin the benchmark portion of this article, but first let me show you our test system.

Mainboard - Gigabyte GA-K8NXP-SLI motherboard, nForce4 SLI, Socket 939

Processor - Athlon 64 4000+, Athlon 64 3800+, Socket 939, Pentium 4 2.8 GHz, 3.4 GHz, 3.6 GHz

Graphics Cards - NVIDIA GeForce 6800 GT (PCX), 256 MB
 

Memory - 512 MB (2x256MB) DDR1 - PC-3200 Crucial 2.5-3-3-7

Software

Windows XP Professional SP2

Splinter Cell (Guru3D custom timedemo)
Half-Life 2