arma 3 poly limit

[aef 10]

Does anyone have rough estimate of what the arma 3 poly limit is? I know arma 1 is 16000 arma 2 is 32000

[nodunit 397]

3200? where did you get that? Whenever I try to load up Arma2 with above 16K it complains about too many vertices and makes it not appear, crashing if I try to start with it..though I've seen a 130K poly unit made using proxies so essentially as long as you use those there is no limit.

[wld427 1705]

3200? where did you get that? Whenever I try to load up Arma2 with above 16K it complains about too many vertices and makes it not appear, crashing if I try to start with it..though I've seen a 130K poly unit made using proxies so essentially as long as you use those there is no limit.

not necessarily.... i distinctly remember Rock telling me there is a limit on the number of proxies that can be used in a single model as well.

[rksl-rock 1301]

not necessarily.... i distinctly remember Rock telling me there is a limit on the number of proxies that can be used in a single model as well.

It depends on the complexity of the proxy. But in the context of that conversation we were talking about cargo proxies. Specifically related to C-130/C-17/A400M paratrooper proxies. We arrived at 32 as the optimal after signifcant SP and MP testing. The performance degraded signifcantly with much more than 32 cargo proxies. Given that the average soldier model is ~5500 faces + weapons in the 0 res LoD. Thats minimum 176000 faces plus the host model and its component proxies.

Simple rule to live by: "The more you add the more you lag." ;)

3200? where did you get that? Whenever I try to load up Arma2 with above 16K it complains about too many vertices and makes it not appear, crashing if I try to start with it..though I've seen a 130K poly unit made using proxies so essentially as long as you use those there is no limit.

32000 was the technical vertex limit given by one of the BIS staff for ArmA2.

Don't confuse Buldozer with the ArmA2OA.exe build. Buldozer is essentially a cut down exe intended for preview use only. It doesn't always accurately represent what you can use ingame.

As for using proxies to increase detail. There is a practical draw limit. You'll find it pretty quickly on a low spec PC. Personally I've never needed to go past 30k faces for any application in ArmA/VBS2/ArmA2/OA. I really don't see why anyone would need to create an ArmA2 model with 130k faces. Unless we're talking about a really detailed multipart carrier etc?

[aef 10]

the unit and i are looking at models on turbo squid and we are trying to find out limit we can put on arma 2. we were looking at models with 60k+ polys but we have the chance to get ones with polys under 35k.

[5LEvEN 11]

the unit and i are looking at models on turbo squid and we are trying to find out limit we can put on arma 2. we were looking at models with 60k+ polys but we have the chance to get ones with polys under 35k.

Although it is much easier to download models then to make them your selfs, the models you download are very high quality, not optimized for games. Thus it is better to make your own...

[nodunit 397]

It depends on the complexity of the proxy. But in the context of that conversation we were talking about cargo proxies. Specifically related to C-130/C-17/A400M paratrooper proxies. We arrived at 32 as the optimal after signifcant SP and MP testing. The performance degraded signifcantly with much more than 32 cargo proxies. Given that the average soldier model is ~5500 faces + weapons in the 0 res LoD. Thats minimum 176000 faces plus the host model and its component proxies.

Simple rule to live by: "The more you add the more you lag." ;)

32000 was the technical vertex limit given by one of the BIS staff for ArmA2.

Don't confuse Buldozer with the ArmA2OA.exe build. Buldozer is essentially a cut down exe intended for preview use only. It doesn't always accurately represent what you can use ingame.

As for using proxies to increase detail. There is a practical draw limit. You'll find it pretty quickly on a low spec PC. Personally I've never needed to go past 30k faces for any application in ArmA/VBS2/ArmA2/OA. I really don't see why anyone would need to create an ArmA2 model with 130k faces. Unless we're talking about a really detailed multipart carrier etc?

The 130K was more or less a gimmick 'can it be done' kind of thing. The error I got was in arrowhead not bulldozer, aggrivating since it was to test out the cockpit which doe not even reach the 30K mark you speak of.

Edited July 13, 2011 by NodUnit

[max power 21]

http://forums.bistudio.com/showpost.php?p=1950430&postcount=4

I think the reason for that is the p3d file format describes hard edges as separate vertices. A soft edged cube would have 8 vertices towards the vertex buffer limit, but a hard edged one has 24. In this case, it seems like there's no efficiency advantage to make a hard shaded edge over bevelling (or chamfering) it.

edit: bad math

Edited July 13, 2011 by Max Power

[nodunit 397]

But wouldn't chamfering just run a big circle since it introduces another layer of vertices?

[max power 21]

Hmmm.. I think it would help to visualize it.

Imagine a soft shaded cube with 8 vertices. Now hard shade them. This is the same as detaching them without moving them, meaning you have 4 vertices per side, with six sides per cube.

Now imagine the faces have smalls gap in between the six faces, and you fill the gaps with soft sided faces. You'll have 12 more quad faces and eight more triangular faces, but no new vertices, the same as chamfering every edge on the cube.

Edited July 13, 2011 by Max Power

[jagheterjan 10]

A cleverly chamfered edge should allow for a single smoothing group across all the bordering faces. Which in theory should slightly reduce the vertex normal count by turning a hard edge into a soft one.

Edited July 13, 2011 by jagheterjan
Max beat me to it.

[dmarkwick 261]

four more triangular faces

eight more triangular faces surely? ;)

[nodunit 397]

So the vertex limit is more important in O2 than polycount? If I were to say take this shape

and chamfer it at that hard edge towards the back, the chamfer would decrease the actual drawcount and would be more efficient?

Even if that were the case though wouldn't the amount of polygons generated from it eventually render the effect moot? Say you take an abrams and chamfer all of it's hard edges on the body, you effectively doubled or more the polycount..is that practical?

Edited July 13, 2011 by NodUnit

[max power 21]

eight more triangular faces surely? ;)

Yep. I was trying to figure out the best way to express it and I was talking about it per corners of quads and I guess was still stuck on that. Of course you all can see what I mean.

[dmarkwick 261]

More like this (exaggerated for effect):

12 extra quads, and 8 extra tris.

*edit*

I seem to have posted that image in answer to a question that no longer exists... I am possibly hallucinating :D

Edited July 13, 2011 by DMarkwick

[max power 21]

Here's a screenshot with a cube soft shaded, bevelled as I described, bevelled with one more division, and hard shaded.

You can't really see it but with you get a nice highlight at the corners with a bevelled edge.

I think the directx face buffer is the same size as the vertex buffer, so a 15 bit buffer allows for 32000 faces and 32000 vertices. I think you usually reach the 32000 vertices limit first.

edit: I should say that I have read some posts on another forum that seem to indicate there is a 'face buffer', but I can't say that I know if there is such a thing.

Edited July 13, 2011 by Max Power

[nodunit 397]

Right, you need two chamfers to help the lighting which is why normal mapping is often used. I had the same issue when working on my apache's EFABS, to get the nice crisp rounding with even lightflow I had to chamfer and then make two edgeloops on the outside.

My biggest concern is how practical it is and if it is the better choice of more polygons vs vertices..not to mention a not so great streak of luck with normal maps in bulldozer for some reason.

Edited July 13, 2011 by NodUnit

[max power 21]

Right, you need two chamfers to help the lighting which is why normal mapping is often used. I had the same issue when working on my apache's EFABS, to get the nice crisp rounding with even lightflow I had to chamfer and then make two edgeloops on the outside.

My biggest concern is how practical it is and if it is the better choice of more polygons vs vertices..not to mention a not so great streak of luck with normal maps in bulldozer for some reason.

Depending on what your mesh looks like, just one can be enough.

[nodunit 397]

Okay I'm going to use my work as an example, since it's a particular chunky sharp shaded part..

separated http://i229.photobucket.com/albums/ee200/NodUnit/Model%20work/AH-64D/Work%20In%20Progress/breakdown1.jpg?t=1307481509

pieced together http://i229.photobucket.com/albums/ee200/NodUnit/Model%20work/AH-64D/Work%20In%20Progress/Apache%20Renders/mast1.jpg?t=1306971218

Are you saying that chamfering some of that would be more productive than hard shading?

Edited July 13, 2011 by NodUnit

[max power 21]

Glancing over the rotorhead, I can't see any place where I would chamfer instead of make a hard edge for a real time model.

I wasn't saying that the chamfering is more efficient, I said that there's no difference between chamfering and hard edges when you're staring down the vertex limit.

[Big Dawg KS 6]

This thread has lost its relevance to ArmA 3...

[-=seany=- 5]

Interesting to read though..well for me it was.

[max power 21]

This thread has lost its relevance to ArmA 3...

You think that none of this info is applicable to modders who want to model for ArmA 3?

[PuFu 4600]

So the vertex limit is more important in O2 than polycount?

vertex normals is what matters, not vertex count, triangle or polygon count (all of it is related in end anyways).

Here is the simplest example, using 2 boxes: The first is hard edged, the second is all smooth edged.

You can easily see that in the case of hard edge box, the number of vertex normals is 3 times the one for smooth edged box.

This is the nth time i am saying it. The O2 p3d 32k limit is related to vertex normals, not vertices, or triangles.

As a conclusion, no matter what sort of modelling path you take (chamfer or not - for HP models, edge loops are far better than chamfers, since it keeps things quads for sub-d modelling),

one could use more geometry as long as he can keep the vertex edges lower than with less geometry :).

In any case, that is why normals maps exists, to deal with the said vertex normals limitations :P

for 3ds max users: http://www.armastack.info/index.php/543/counting-the-vertex-normals-when-working-in-3ds-max

Edited July 14, 2011 by PuFu

[Big Dawg KS 6]

You think that none of this info is applicable to modders who want to model for ArmA 3?

It's not specific to ArmA 3 though. It seems more appropriate in Offtopic or maybe ArmA 2 editing thread.