Uday,
When simulating requests, LoadComplete uses the data (user name and password) it saved to the user scenario during the recording.
So, if you want to simulate multiple virtual users that log in with the same credentials, you just need to simulate your "login page" scenario with the needed number of virtual users in your test :-).
To determine the number of VUs that will cause a crash, run the test with various numbers of virtual users. Or, use the Stepwise and Continuous Load features to increase the load dynamically. See Load Testing: Checking Whether Server Supports Certain Number of Users.
Also, I'd suggest that you define what "crash" will mean in your case. It's quite possible that the server will remain functioning even under heavy load, but the response time will be too long, or, for example, the server will respond with codes 500, 503 or 504 rather than 200.
So, perhaps, you should wait not for a crash, but for long page load time or different response codes.
You can define the target load time on the "Quality of Service" tab of the test editor, or in the LoadTesting editor. See http://support.smartbear.com/viewarticle/24829/. When the "Quality of Service" settings are violated, TestComplete posts an error message to the log. You can look at Runtime Graphs to determine the number of VUs at which an error occurs.
If the response code for a request differs from the recorded code, LoadComplete will post a warning to the log. You can look at Runtime Graphs to determine what number of VUs caused the issue.