>>I think the second should be faster, so I would be surprised if you got different results in your tests.
>
>What's a good way to test so as to eliminate the effects of cached data and stuff? The data is on free tables on a server on our lan, and I am runnnig the app on my desktop, so the data is coming over the LAN.
>
>I want to be sure I do not mess up the results by having it the data cached or something.

Run them alternately, one after another, in a loop. So if one gets the benefit of caching, so will the other.

You can also apply a net limiter (it's, I guess, a freebie program you can download somewhere on the net, try SourceForge or whassitsname) and squeeze your network pipes with it - that would make the effects of server side caching less pronounced and show the difference in more contrast. Or you can do a brute force clearing of the cache by rebooting the server between tries, or by disabling the cache there altogether, but I guess that's exactly brutal.