>The database in question is only 50 MB on disk and the tables I'm talking to are generally only a few hundreds rows and maybe 20
>columns nothing but char, int and float. It's all very small and light!
>
>Running on the server, over the VPN and using RDP, the simple queries I run complete almost instantly, in hundredths of a second.
>
>Running on my client, over the VPN using ODBC directly, the exact same query takes about three seconds.

Exactly what is the definition of your fields? I would expect (or rather, would be slightly surprised if not - I really don't know anything about transferring things from the server to the client) a CHAR(1) field to be considerably quicker than a CHAR(32000) field - even if both of them just contained an A (followed by a lot of spaces in the CHAR(32000) case).

And what does your select look like (just out of curiosity)?

Set