When in the process of improving our software / os performance, there are often tweak guides mentioning to turn on /off certain features. However most times there are no benchmarking graphs that support the claim of performance increases. I'm thinking of things like turning of windows services, startup delayers, effects of fixed pagefile sizes etc.
Therefore I was wondering if anyone know of the best tool to run performance tests. Ideally it would allow you to run the right tests for multiple times, compares results with other runs (for the before/after effect) and outputs comparative results + graphs. Optionally these last steps can be done with something like Excel or another spreadsheet program.
I know Bootvis would be a likely candidate: it measures windows startup times and shows graphs of driver/service delays etc. However it is no longer supported on XP and it will cause problems on Vista. Also there are probably other tests you'd want to run, not just boot up times. A good tool would probably contain a better set of tests. Advice on which tests would measure 'performance increase' would be appreciated.
There are also performance testing tools within Vista, but they focus on making sure the system is running ok. It would need an additional layer to explain the results and how they impact on the user experience.
Then lastly there are general benchmarking suites such as 3dMark (for video) Sandra Sisoft but they again provide too much info and not distilled into a easy to compare format for system / software tweaking.
Any ideas, discussion and suggestions would be great.