I have been looking at a lot of videos of people showing off distruction in engines like UDK and Blender. I noticed that each time the objects were broken into smaller pieces the frame rate drops dramatically. I then also noticed other videos with much more realistic destruction but with much smoother frame rates although I did notice in the more detailed videos the camera positions was fixed. I can't really understand why a game unlike a cinematic like video has so much less detail, its not like polygon count is an issue, would it be to do with the physics that apply to each block as it is broken and detached?
Why can you get smooth frame rate like this....
but not when you actually play as a character like this....
In the first video, the environment is simple. There is nothing in the world, except the objects that are being acted on by physics. This would not work well in a real in-game environment, because along with the physics there would be AI calculations, lighting, shading, and many other processes going on at the same time. This would make it too slow to run in real time.
If the objects break up into smaller pieces, it will mean there will be more pieces, and more pieces means slower to calculate. In each frame of a physics simulation, EACH SHAPE has to be checked for collision with EVERY OTHER SHAPE (with some optimisation techniques). If there is a collision between two shapes then the new velocities have to be resolved. The more objects you have in memory at one time, the longer this process will take.
The first video you posted is an animation I believe, so I'm not sure if it was rendered in real time. That would explain why the frame rate is smoother.
Really good question by the way, I love stuff like this.