mesh sizes and texuring...hmmm

[luckymutt]

I've alluded to this in a couple of past posts, but lemme spell it out here.
(this will prolly come down to opinion and preference)
There are some levels that I am envisioning and starting to model that, if I do it the way I am used to, will wind up being mind-bogglingly huge, so I am trying to find ways to be as efficient as possible.
For example, should I:
a) tile a texture on a surface to somthing like, say, 20x20,
b) used photoshep to make the image equivalent to the 20x20 then not use any tiling and just apply it
or at the end of the day would it not make any difference?

also, in order to conserve geometry, I'm warming up to the idea of modelling with planes more that with boxes. That makes it a real pain in the modelling end, but considering the poly count I would be saving, I'm thinking it would be worth it.
I wanna squeeze as much detail out as possible, and using bump maps just isn't going to cut it.

What is the impact difference between using something like MY3D with the light mapping (and larger file size) versus using a 3DS file (or whatever else) with dynamic lighting?
Again, is the bigger file with lightmapping balance out with the need to add dynamic lights?

So basically, I suppose, what short cuts has anyone else found to get something to look GOOD while keeping things as resource friendly as possible?

[disanti]

Really the only difference on performance is going to be loading time. If you load a 3ds file that has the exact geometry to a my3d file, they are going to look the same in memory. Only difference is that the my3d file has lightmapping too, that is what makes the size bigger. If you gave the 3ds file lightmapping, it would have the exact same performance. Really, I would recommend using my3d instead of 3ds in irrlicht... 3ds seems to have a lot of bugs right now.

If you are truelly worried about my3d's size, you can try out the directx format (.x). When you compress it into a zip file it is pretty much only a few kb, but no matter what, when you load it it will be the same size in memory as a 3ds or my3d file that has the same geometry. If you do choose to use .X, remember NEVER put more then one object in it if you want irrlicht to be stable.

Hope this helps~ disanti
________
Ferrari daytona

[luckymutt]

If you do choose to use .X, remember NEVER put more then one object in it if you want irrlicht to be stable.

Is this why when I try an X file with multiple objects they all centern around 0, 0, 0? All on top of each other?
I have no problem loading them seperatly, but would they retain their realative positions that way?
That has been my only problem with the X format.

So far, the 3ds format (exported from max) doesn't seem to give me any problems, except for getting any decent lighting on them through Irrlicht.

What about poly count? If I understand correctly, with the OctTree it only needs to calculate what is actually visible at any given time, but what would a reasonable count be for what is visible at any one time?
Certainly that would depend on the user's hardware, but what do you suppose would be "average"?