[mikhael]

What are we looking at for a reasonable performance for higher poly ships?  I got a little carried away with a model and its hit about 9000 polys. I think I'm going to hit 12k all up, with all sub-objects.

I may need to have two or three of these in a mission. What can I expect in the way of performance on a reasonable system with three of these and about 12 to 20 fighters on screen?

[aldo_14]

Try Omniscapers as a benchmark, maybe?

[Nuke]

i just finished a 2000 poly bomber.

[aldo_14]

congratulations

[Bobboau]

I think that'll be fine, the perfomence of rendering a relitivly small number of ships is equivilent to most modern era engines, so it'll probly handle it as well as freelancer's engine.
as seen in omniscapers ST conversions we can handle a rather large number of polys for a rather large number of ships, only thing I'd sudgest is try to use less than 8 textures on any gien subobject (I don't expect that to be a problem for you but I'm mentioning it for others)

[Taristin]

I had decent framerates with my 14k poly Anuket. And with the Ryx ships as well, and I have a crappy GF2 MX PCI... so...

[StratComm]

I have noticed that a lot of the slowdown characteristic of the "later builds" is actually related to textures.  There are a couple of threads about this already, but basically I think Lightspeed's stuff needs to be resized (all of it) to 2^x dimensions rather than the current oddball sizes they are using.  The game does this anyway, it'll just take some of the on-the-fly calculations out.

[mikhael]

We'll see how it goes. I'll probably get someone on the 158th to pof it without textures and put three or four of them into a furball mission and see what happens.

[Lightspeed]

Originally posted by StratComm
I have noticed that a lot of the slowdown characteristic of the "later builds" is actually related to textures.  There are a couple of threads about this already, but basically I think Lightspeed's stuff needs to be resized (all of it) to 2^x dimensions rather than the current oddball sizes they are using.  The game does this anyway, it'll just take some of the on-the-fly calculations out.

Some of it, yes.

I'll wait with that till DDS is properly implemented and tested though (with working alpha transparency on planets!) so I can get an even better performance increase.

[KARMA]

btw mikhael, you can reduce pcount by splitting mesh in submodels and removing the hidden faces. If grouped together PCS will convert em as a single subobject.
I know it's a weird thing to do usually, but it can be useful when you need to cut some thousands polys without reducing details.
Since you can't test directly, you should avoid intersections (wich may or may not work), but if you don't use stupizilion bilions of submodels you should be safe, and reduce a lot the pcount.

[Bobboau]

you realy should put some texture on it, I don't think non-textured pollies were ever suported

[Nuke]

when modeling under truespace, more polys (and textures) means more work. my impatience tends to keep most of my models under 3000 polys.

using truspace creates of errors. i imported my apis model into max4. using certian modifiers i managed to eleminate about 100 vertices that aperantly didnt connect to anything. disturbed by this, i loaded several other models and found similar errors. i really need to figure out how to use max, because truespace really pisses me off.

[Bobboau]

ugh, those errors are probly made by your conversion, not TS, a lot of them breack them up into composit triangles each with there own set of verts

[Nuke]

i tried several formats and got the same errors.