| Message |
Not with cars, but with motherboards, video cards, and CPU's. I used to be the tech at a MoBo manufacturer. I can tell you with 100% confidence that benchmarking leaves tons of room for interpretation. It can be a shady practice. Just about when any company is asked to provide a sample for a benchmark comparison, you can be almost certain that the "generic" sample you get is a finely tuned specimen made just for that purpose. In fact, some companies would write different firmware that would excel in each specific benchmark, such as one version using more aggressive timings on the memory subsystem to make the mem throughput seem higher, then another version of bios would slightly overclock the PCI bus to make the hard disk throughput seem higher. You couldn't have one version of bios that ran everything aggressive though, or the board would be unstable. Of the companies I have tested, Intel is by far the dirtiest. They'd go as far as pressure you to include certain features to benchmark for. These features could be absolutely worthless, but they are features that their products had and their competitor's didn't. They even would pressure the maker of the benchmark to change the way the benchmark worked to show their products in a more favorable light. Look at WinBench... one year AMD's chips dominate, next thing you know Intel is courting ZDLabs and sure enough the next version has compatibility problems with AMD's chips, crippling their results. Coincidence? No way in hell. What I'm getting at is if someone is going to benchmark something, the methodology should be legit first.
 |
 |
