How long until you lag?

This is for those who don't believe that mass numbers of vector-controlled objects are feasable performance wise. I've heard it said that any more than 50 will lag!

This map will keep spawning 10 objects per second. Collision is disabled and they are rendered at 30 frames per second. This could easily be faster if I stripped out calculations that are not used in this demo like forces, friction, and the improved integration method. But this is to show how fast it can be under in-game circumstances with the calculations for all those features enabled.

Post how many objects your computer handles before the framerate starts to decrease noticably!

I made it to 500 before it started to slow and 600 before it froze.

Dropped below 30 FPS at around 350. 15 FPS at 500.

This is with half of my processing capacity being used up someone else logged on, and lots of other programs running.

Same result here. Started to slow a bit around 480, tho.

Starting noticing performance hits around 250; lag at 400; quit at 600 where frame rate dropped below .3 fps

Keep in mind, I'm running all specs at high and 1024x768x32

Will lower specs and try again.

Everything at low (same res) performance hits at 380; lag at 440; quit at 600 (frame rates weren't as bad as before.)

I just tried it with no graphic model for the marbles and I got up to about 540 before it started lagging. Not too much of a difference.

It was at around 500 when i started lagging for me as well.

15 fps at 250, 10 fps at 300 and I didn't really check after that. But my computer semi-fails (only 512 ram :< and I'm just waiting to get a laptop so I can't be bothered to buy more).

Seems collision objects, on this map, have about half the performance. I started lagging at 250 objects. But I also turned on sounds for bouncing and collision, and I had the collision realism options set to high, which decreases performance considerably.

The performance cost of making an object collidable is directly proportional to the square of its speed limit (and thus how large the area of the unitentersrange trigger is) and how close that object is to other objects (and thus how many objects are likely to be in the unitentersrange area). Using objects that don't move as fast and spreading them out greatly improves performance.

So typically a map using more than 250 collidable objects would be much larger than this little demo map, and the objects would be spread out farther apart, and it's not even necessary to make them into "objects" until those objects come into use by players anyway.

So assuming that collidable objects cost 1x-2x as much performance on average as non-collidable objects (depending on if they are engaged in lots of collisions or none) and a map has a 1:1 ratio between collision and non-collision objects, you'd expect to be able to use somewhere between 333-500 objects at once "under real game conditions" with high realism options before lag begins.

Of course that's on my computer, so if you got a worse score scale that range down proportionally.

I'm curious, can anyone imagine a potential use for the engine where those performance limitations would ever become a problem?

A TD thats all about killing falling balls. But even with such a thing, balanced correctly, it could run decently on most computers.

I think this used on a TD would be overkill

No way, I have had an idea about a TD based on using physics effects to keep units at bay instead of mazes for a long time. Who doesn't want a wind tower that knocks everything back, or an earthquake tower that knocks things up and down...