Fireball casting speed

100 int = -2% cast time? that doesn't sound right. Check my experiments, it's the thread titled "Experiments on Fireball Cast Speed".
You'll see that there are a few data points that relate Int to cast time.
My hypothesis is that it actually follows a logarithmic curve, since 3 seconds are dropped going from 0 int to about 500 int, another 2 are dropped going from 500 to 1000, and 1 second is dropped going from 1000 to 1500. I also have a hunch that the mp cost reduction is simply due to the reduction in cast time, the mp drain over time as one casts stays the same.

Edit: It's not logarithmic apparently. The equation is Cast Time = Base Cast Time/(1 + Int/4000)

Now if you want the absolute fastest I could push the cast time to, I got 7.448 seconds. This is running through the calculations with my 18 second base cast time. I could probably push it to about 7 seconds at 1500 int.
As for most efficient, I have no idea what you're talking about.