Quote:
Originally Posted by JKyle
The reason I did the testing in a casual way is because I wanted to see how much performance improvement I could get from the new build on everyday use.
|
Indexing a job twice in a row with different versions is
not everyday use! The first run may not benefit from the caches whereas the second always will. That is why I recommend to purge windows and disk caches before every run when doing performance testing. So if you would run the first time with the new version and the second with the old version, you won't see the same gains as you would if each ran without caches.
Alternatively, I suppose you could do it three times, using the first to just fill the caches and ignore its result. Then both following runs would benefit from the caches.