Testing Methodology

First off, for this look at the effects of CPU speed on gaming performance, we decided to stick solely with the Core 2 Duo as our CPU of choice. While probably many of the PC gaming public still has single-core machines -- for example, the latest hardware survey performed by Valve / Steam shows that just under 62% of their players have single core CPUs -- we thought that this type of article would be of most interest to people who are currently, or sometime in the near future, upgrading their systems, and are wondering what kind of horsepower they need. By the same token, we decided not to go the other way and look at very high end CPUs, such as a Penryn quad-core processors as not only are they currently priced out of reach for most, but in  our experience four cores do not generally offer much of a performance increases over two cores for gaming.

If you are looking at going to a Core 2 Duo system for the first time, you can get one for as little as $100. If you wanted to spend a bit more, you could get something like a new Wolfdale 3GHz Core 2 Duo E8400, for not much more than double that hundred dollars -- Core 2 Duos prices are falling fast, so it doesn't take much cash to get your hands on a tremendous amount of processing power these days.

For these reasons, we decided to go with an E6700 for our testing. The E6700 is somewhere close to the middle-of-the-road as the Core 2 Duos go, and is capable of  scaling to wide range of speeds. Starting testing at the E6700 underclocked to 1.6 GHz, we decided to test performance gains for every additional 200 MHz, while trying to keep the memory speeds as close to 900 MHz as system stability would allow.

With speeds ranging 1.6 GHz to 3.0 GHz, we thought we'd get a good indication of how much of an effect that an E4300 would, say, compare to a E6850, when it came down to frames per second. Even though it would make sense to run our benchmarks at a very low resolution, such as 800x600, to gauge CPU scaling performance, we wanted to use resolutions that were much more likely to be used, so we went with 1280x1024 and 1600x1200.