For many of today's "fanatics" (to quote Michael Dell), the ability to overclock a CPU and video card is crucial in their decision-making process. I decided to use 3D Mark 2000 to demonstrate the effects of overclocking. It provided some interesting results.

I was not able to overclock the memory on the Erazor X2 consistantly enough to warrant inclusion here. The heatsinks I mentioned earlier would definitely help that. However I did include, for comparison purposes, the overclocking of the core of the Erazor X2 from 120MHz to 125MHz. [side note: this card was not pushed to its limits] I've also included results of overclocking the CPU. As you can see, overclocking the CPU brings nice results, though for a 20% increase in CPU speed, there was only a corresponding increase of 14% in the 3D Marks. A 4% increase in the Erazor X2 core only brought a 2% overall performance increase...probably not worth the extra heat generated by the chip. The numbers that are surprising are the results for the High Polygon Counts. nVidia has been criticized for the poor Lighting performance of the GeForce's T&L engine when multiple light sources are used. However, look at the results of the multiple lights when the FSB is increased to 120MHz. This yields an AGP speed of 80MHz (AGP spec is 66MHz). Now for a 20% increase in bus speed, we see a 35% increase in 4 lights and a whopping 48% increase with 8 lights! It looks like the AGP bus is a bottleneck when it comes to Lighting throughput. I don't have an AGP 4X motherboard, but it would be interesting to see if those kind of results hold true for a non-overclocked system running at AGP 4X.