The issue you describe is a typical lack of RAM memory issues. You can try to improve your script in terms f memory usage, meaning loading one table a time, storing in qvd and drop the table afterwards. Once you have all the qvds, try to load them in a consolidated model, if this does not work for you means that you will have to buy some more RAM memory for the server.
the amount of load is independent of your source. Wether RDBMS or textfiles isn't important. QV works fine with very redundant data, in this case the compress-rate may be very high. Do you need every field etc. ?
Ivan told you some good workarounds but I think you will need sooner or later a machine with a 64bit-Windows and the according RAM. Your data is growing fast ervery month.