Quake3World

I don't know why, but my hint brushes doesn't seem to work, not with hint/skip and not with all faces as hint either. I've used them before, but this time seems they don't work. There are not leaks, and the VIS works for other parts, but I can't see that where I've placed hints there are splits. Why? I don't remember having to use any special command when I compile to use the hint/skip feature.

And also, when I did prayers nest 2 a couple years ago, I was told that a Tri-count on 10K should be the "max", I don't know if that was true, but if so, what is it today? on a mid-range system? And, what does the second tricounter says in the r_speeds 1 counter, there are two of them separated by a "/" and they doesnt say the same thing even though the difference is small.

It was 6k for VQ3 and 10k for TA, they're still valid as limits regardless imho

6K sounds a bit low.. but ok.

Anyway, found that q3map2 won't produce a .prt file, I guess that shows that there's a problem somewhere... ?

Uh, there's no prt generated from any map really, well, not my VIS-test map either, which basicly is just two simple rooms and a hallway... hmm

6k tris is silly for a limit these days, sorry Kat. My buddy has a mid-range system and he gets a constant 350-400fps on every Quake3 map running max settings at 1024*768. 350fps is overly excessive, so I wouldn't worry about doubling or even tripling the tris count. Just watch what kinda shaders you have and what they're doing. Tris aren't the only thing that can kill the FPS... multiple pass shaders can hurt them pretty bad too.

I'm working on a map right now and the main room is about 22-23k tris. Even on my shitbox 5 year old computer I get about 30fps on it, and this is on the same settings mentioned above. If I turn it down to 800*600 and lower the quality of everything to mid-way, I can get up to about 80fps which is pretty good considering my machine is a pile of crap.

Hipshot wrote:6K sounds a bit low.. but ok.

Anyway, found that q3map2 won't produce a .prt file, I guess that shows that there's a problem somewhere... ?

Uh, there's no prt generated from any map really, well, not my VIS-test map either, which basicly is just two simple rooms and a hallway... hmm

Yeah if you aren't getting a .prt file I think that's a problem. AFAIK the .prt file is basically all the portals that VIS uses for calculating visability. If you have no .prt file, you have no portals, and you cannot calculate vis. What do your q3map2 switches look like?

mrd wrote:

Hipshot wrote:6K sounds a bit low.. but ok.

Anyway, found that q3map2 won't produce a .prt file, I guess that shows that there's a problem somewhere... ?

Uh, there's no prt generated from any map really, well, not my VIS-test map either, which basicly is just two simple rooms and a hallway... hmm

Yeah if you aren't getting a .prt file I think that's a problem. AFAIK the .prt file is basically all the portals that VIS uses for calculating visability. If you have no .prt file, you have no portals, and you cannot calculate vis. What do your q3map2 switches look like?

Fond a command, "-saveprt". And now I have one... strange though, I have never used this before and I have many PRTs in my maps folder from older maps... was this default in older versions of Q3map and Q3map2?

The .prt file is deleted by the vis stage after it's done with it so your maps folder doesn't fill with junk.

So if you run full compiles you won't ever see one without adding -saveprt, but if you're running quick bsp -meta compiles to test for leaks etc. you'll see them left lying around.

A few years ago, the top tris amount was 20K (pushing the limit though).
I found out that at the time I couldn't play my map with a steady 125 fps (full detail, 1024*768).
Since then, PC hardware has moved on quite a bit, and maps that were pushing the system (with comments as "eats fps" etc) are now easily getting 200+ fps on same high settings.

My map has around 42k and plays very very nice on on older card (fx5950). It's easy to test too, you can run several other maps on your system and see what triscount you get, and extra/interpolate the tris for a "general" PC, depending on what system you have. I think 30k is no problem anymore, but as said above keep an eye on the shaders as well.

mrd wrote:6k tris is silly for a limit these days, sorry Kat...

Learn much you do have to young padawan

If you can describe the major features of a map section under 6K tris it means you have a solid fundamental structure in place upon which you can then add more detailing, it also means that because you *have* done that you can go crazy with effects and shader shinanigans.

The underlying message there is not to be waistful or messy with your brushmonkeyisms just becasue you can load 100k tris into a Q3 scene

True, just because your hardware can draw the entire map simultaneously at decent frame rates doesn't mean that you shouldn't have the need to properly think about optimization.

I agree totally, that's not really what I was getting at. Personally when I make maps with tris counts that high, it's usually because I'm doing a lot of intricate brush work, not because I'm getting slack with hints or caulking stuff.

Does the Quake3 engine itself have a limit to how many tris can be displayed?

Quack wrote:Does the Quake3 engine itself have a limit to how many tris can be displayed?

Probably, but I think it's so high that it's hard to reach...

Quack wrote:Does the Quake3 engine itself have a limit to how many tris can be displayed?

Yes. 65,536.

What happens after that? Missing textures, no shaders, parts you can see through...?

It wraps around (e.g. #65,537 becomes #1). The exact behaviour will be a tad undefined.

There is something i misunderstand, i reached a lot higher value. without any trouble. (except pretty low framerate).

Yeah, I'm pretty sure I've seen 100k tris counts in r_speeds before...

Much more

[lvlshot]http://witistyrannic.free.fr/Curvy12.jpg[/lvlshot]

It seems you're right. I checked the source and there indeed doesn't appear to be a surface limit, the engine just keeps outputting triangle strips. A bit particular, because I remember bumping into the 65,536 limit I mentioned before but I can't find any trace of it now. My apologies.

EDIT: found it. 65,536 is the max amount of verts bspc can handle.

Weird, what source ? (not that i don't believe you, it's just that i am curious).


I remember that i reached the more than 1M tris without troubles, then around 1.3M tris i had to modify the com_hunkmeg variable or i wasn't able to launch the map.

near 2M tris the map wasn't playable (all texture were white or grey, no lightmap, ...)

^misantropia^ wrote:EDIT: found it. 65,536 is the max amount of verts bspc can handle.

BSPC? Wouldn't that imply the max number of bot collision surfaces instead?

Ho bspc ! i thought it was bsp.

That way i perfectly understand, i wasn't able to compile my map for aas, except with a serious botclipping.

Are there limits for the viewable distance of an object? Let's say you have a skybox, and the distance from the building you are standing in and the structural object being rendered is about 30,000 units away. Since the skybox has a very low tris count, could you add some very low tris structural to add perspective without any serious rendering problems? Most of the things I would like to add are just simple planar images of buildings/motifs that mesh in with the skybox.

You can do it, sure, but if it's too far away you may get some z-buffer issues. 32 bit z-buffers only work so far.