A have amd 3000+ 64, GeForce 6800GS and 2Gb DDR 400. When there are a lot of kicks or punches, so that figures touch one another fps drops to 20-30. Is it ok? I tried this game on another PC with P4 1.7 fnd GeForce 6600GT and had about 10 fps sometimes. Is it normal and all people have such a problem?

im not too fancy with computers but i seem to run around 20-30 fps on a laptop and this seems to me like the game was designed to run in this setting or is it supposed to run faster?

Yeah, this is because there are lots of physics calculations to do so it really hammers the CPU.

6800GS probably, i heard it's extremely buggy, unless you typoed and ment 7800GS+...

this is normal, yes

6800GS is not buggy and why why the developers of these great game cannot make these calculation not so difficult that we could have 60 FPS all the time?

Use the search feature of the forum.
http://www.toribash.com/forum/index.php?topic=205.0

I have a question: the game calculates the position of parts of characters very exactly, for example 0.0004555 and this is the main difficulty for the processor and slows down fps, but is it necessary?

It will take longer if you have to start rounding off numbers.

The FPU (floating point unit) will take the same amount of time no matter how precise it it. Also, it is not all the accurate, the FPU does take a bit of a guess at times. This inaccuracy is usually small, so it can be ignored.

The FPU will take (for description purposes) 1 clock cycle to add. If you want to round first, you have to take at least 2 clock cycles for the rounding of each value.
e.g.
for 0.00455:
(1) is number greater than 0.0045?
(2) if not, 0.004, otherwise 0.005.

I think the problem comes from the number of calculation that needs to be done not the accuracy of the numbers. Because if both players are bunched up close together a large number of body parts have to be tested for collisions and forces against a large number of other body parts.
So the number or calculations kind of goes up exponentially. I think the equation for the number of collision tests to do is something like x*(x-1)/2 where x is the number of objects (of course there are ways of optimising like grid subdivision). But that doesnт€™t even take into account the calculations it takes to actually find if the objects are colliding, then once you find a colliding object you need to work out the forces to apply then you have to find out if the new force will effect another object. So as you can see itт€™s not a simple procedure.

Well thatт€™s what I think with my limited knowledge of collision detection/response.

Anyway thatт€™s enough of a physics rant from me.

Fatty out.