I assume you're running it in QB64...

My results were more consistent in QBasic/DOSBOx, but not in QB64. The reason is simple. During the loop that measures the speed of the user's computer, there is a test with each iteration to see if TIME$ has incremented. There is no such test in the loop that makes use of the value thus generated -- the delay loop. As a result, the delay loop executes more rapidly than the test loop, and in QB64, almost immediately. I would suggest using a test of some sort in the delay loop for more consistency.

In any event, this is the type of discussion that we should have been having about Minty's code, not merely dismissing it as foolishness.

-Bob