Tuesday, April 20, 2010

How Benchmarks Don't Stack Up

An interesting coincidence occurred today. AMD posted an article about how synthetic benchmarks don't represent the real world. This is yet another example, one a little more scientific than my previous posts on rethinking your choice for computer hardware (Intro, CPU, Motherboard).

In my prior posts I never came out and stated it so clearly, but the message was implied that simple metrics like CPU speed, FSB speed, or in this case benchmark results, don't necessarily paint an accurate picture on the overall system level performance of your new computer rig. What is most important is that, when selecting a new computer or computer components for do-it-yourselfers, you research each component and how well they communicate with each other.

Just like there are bottlenecks within networks, there are bottlenecks within computers. Loosing 2% performance on a single component may make zero difference to the system of a bottleneck exists elsewhere. Save the cost of the component and spend that money on improving the performance of the bottleneck. How do you find your bottleneck? That's a catch-22, because you need to run benchmarks in order to find it.

Here is a list of common benchmarks, care of Wikipedia.
There are plenty of others. It doesn't take an internet genius to do a Google search to find them.