Couple of quick points you can look at:
Are you able to use a 64bit driver or 32bit?
Newer ODBC driver?
Maybe an OLE DB driver would be quicker?
How are you loading the data, is there alot of manipulation or is that just a straight table load
Any performance problems of the database server
Connection speed to the database server.
As you say 200,000 rows in 45 mins is quite slow.
The load speed is dependent on the load on the server, the speed of the driver and the load on the network . You would need to analyse these to see what is limiting the spreed.
Unless you are doing something really unusual, 200,000 rows is a small data set and it is probably not your machine or load script causing the slow load.
What is the ODBC data source? Non-normalised data sources like Universe can be very slow over ODBC as the ODBC driver has to normalise the data first.
As a last resort:
Keep an eye on the script execution window, is the row count extracted each time increasing in chunks exactly the same size each time after a delay?
If so ask you DBA what the Session data unit (SDU) is on the DB (default 8k 11gr2 Oracle) then try running the command 'netsh interface ipv4 show subinterfaces' on the QlikView server and make sure the Maximum Transmission Unit (MTU) is below this value. I have investigated this before and ended up tweaking our driver settings to allow more throughput to QlikView.
I am having a similar problem, and decided to change the fetch buffer size. However part way through the run the setting reverts. Are you aware of any other way that this setting could be changed, or do I need to change it in a different file, or even the loadscript itself?
Let me know if you need any other details about my system.