A widespread problem in the computer world right now is the tendency to run benchmarks to determine the superior CPU/VideoCard/etc. The whole notion of what benchmarking is has been lost on the online community and many flame wars have started based on incorrect perception of what benchmarking really is.

Benchmarks cannot be used as an accurate basis of comparison between Macs and PCs. Benchmarks test only how fast a program runs on a specific machine and nothing more. Many people think that because Platform A runs Benchmark A faster than Platform B, then Platform A is in fact faster. The only answer that a benchmark gives you, is that Platform A runs faster than Platform B on this current program. A benchmark can easily be changed to show one platform in a better light than another platform. An example of this is Intel's investment in the benchmarking firm BAPco. In their investment, they hope to get benchmarks that show that their CPUs are the fastest CPUs on the market today. However, just because Intel's processors score higher than other processors in BAPco does not mean that Intel processors are faster, it just means that Intel's processors are faster at running BAPco benchmarks.

This leads many people to judge various benchmarks as unfair because they feel that it gives an advantage to one processor over another. However, optimizations are an important part of the development process. If the software developer knows that the majority of his clients are using a specific hardware setup, then it would be logical for him to optimize it for the hardware setup that the majority of his users would be using. Therefore a benchmark optimized for whatever platform he was developing would give a better picture as to how fast his programs would run comparatively on the different platforms.

In an attempt to alleviate this problem, companies have come up with real world benchmarks as opposed to synthetic benchmarks. These benchmarks run tests based on real world code, code that has been optimized to run as fast as possible on a wide array of different machines. However, the tendency is to run one benchmark and come up with a statement such as, "Since this program is floating point intensive and runs faster on Platform A, Platform A has a better floating point unit." However, due to the scenario presented before, optimizations may have favored one platform over another and perhaps another program may run floating point calculations faster on the other platform.

When Steve Jobs presents benchmarking information that shows the Mac is faster than PC, or when PC enthusiasts present benchmarking data that shows otherwise, the reader should be wary as to what they are testing. In the majority of the benchmarks, the PC users are presenting data that is irrelevant to the Mac user, and the Jobs is presenting information that is irrelevant to the PC world. For example, a graphical artist should not be concerned as to how fast a PC can run Quake 3 vs. a Mac but rather how fast the programs he uses such as Photoshop or Illustrator perform specific tasks run on each platform. Likewise a gamer should not care how fast Photoshop runs on his machine compared to another platform, but rather how fast it runs his favorite games. If the point of showing benchmarking data is to convert users from one platform to another then, both parties are failing miserably because of their failure to present relevant benchmarks.

The obsession with speed has clouded over the primary reason people get one type of computer over another. Since Mac apps can't easily be recompiled to run on a PC, and vice versa. People want to be able to communicate easily with their peers and usually the easiest way to do that is to run compatible platforms. Comparing speed between the two platforms is not a very useful thing when it comes to picking a platform. Until a Unix type arrangement has been made where both platforms have identical file systems and applications can be ported over easily by the end user, the comparison in speed is useless.