Thompson M1A1 - Subdivision question

useless99

Hey there!

I'm an absolute newbie to the whole 3D-modelling thing and have just recently started some meshes after I got 3DS Max and VRay. I'm mainly modelling firearms.

Now, my question is, I have been discouraged from using Subdivision (turbosmooth / meshsmooth) for my modelling work, since the final mesh will not be for in-game use but rather for hi-def shots with ray-tracing materials (VRay). I was told that having lots and lots of unnecessary polys will drastically increase render time. Therefore, smoothing edges should be done mainly by using chamfer etc. in order not to unnecessarily increase render time.

Now, while skimming the threads here and on several other forums, I noticed that most high-poly gun models do have smoothed edges even where you would not expect it, i.e. where you would have a completely straight edge when not using sub-d. This adds a lot of neat-looking highlights to those rounded edges, giving more "depth" to the model. However, those high-poly models oftentimes seem to have an almost impractically high poly count (i.e. some high-poly guns I've seen on this forum are at 800'000+ polys!!!).

While I do not need to keep the poly count as low as for game models, I've been able to model my guns with rather low poly count. For example, for this WIP Thompson M1A1 here, the tri count is around 8'500, i.e. around 4'000 polys:


In order to make some of the rather harsh straight edges more rounded, what would you suggest? Does a REALLY high poly count hit render time that hard with raytracing etc. enabled? Would you suggest using sub-d and then manually removing geometry? Or do as I did up to now, i.e. create the mesh and then add rounded edges manually? This would most probably mean that I'll have to chamfer a LOT more edges on that receiver pic posted above. However, it would also mean that I can have rounded edges everywhere without having a ridiculously high poly count by having the larger flat polys subdivided, since no sub-d is necessary there.

Frankly, I'm not too keen on subdividing the model to the point where the viewport in 3DS max slows down noticeably. However, I can understand that this may not be a concern to people who only want to use the high-poly model for texture baking and then discard the high-poly model.

Any input would be greatly appreciated!

Thanks a lot!
useless

gsokol

When people make guns with such large polycounts with subdividing and whatnot, typically they are preparing the high poly version for normal map baking onto the low.

Grimm_Wrecking

Racer's Tut's

Racer did a great job on his tut's. His ak tut should demonstrate more clearly why people are doing 800,000+ poly guns.

But gsokol really summed it up.

Kitteh

Typical gun models are 2000-6000 triangles, and use normal models baked from a high polygon with hundreds of thousands of triangles.

Ghostscape

If you're rendering in V-Ray then why not just chamfer the edges you need to chamfer? add a couple of edges to the fillet rather than the default and you'll get a nice rounded edge. It causes some nasty geo IF YOU WERE GOING TO SUBDIVIDE, but you're not, so who gives a fuck?

One of the big reasons for the sub-d push for normal mapping is it gives you a very smooth curve on all of your edges, and only curvature (not depth) is picked up in a normal, so getting that curvature nicely represented is important. For a still/animation rendered offline (like in V-Ray) then you don't need that kind of curvature, just a bevel or two, usually. Depending on where the camera is you don't need to worry about most of those edges, anyways.

Onee more thing to consider - for normal map baking, typically the edge softening is exaggerated since your curves can't be thinner than a pixel (which might mean your sharpest curve on a environment mesh might be a centimeter wide or so). With a high-rez still you don't need to exaggerate as much to capture those edges on a low-res texture, so you can get away with making them much smaller and more realistic. Many machined edges, like on that gun, may in reality be exceptionally sharp, but would be exagerrated for an in-game model to help pop those edges more often for illustrative/stylistic purposes.

PredatorGSR

I've only done game stuff, but would film follow a similar workflow for creating displacement maps? I.E million poly gun baked down to a 50k with displacement maps?

useless99

Thanks for the replies guys.

gsokol: Yeah, I got that (see my post), but I don't see the need to do this on my model. I was just wondering whether the same results (i.e. mainly the rounded edges) could be achieved without subdivision, and how to go about it.

Grimm_Wrecking: Thanks, I saw the tutorials, but as said, I'm not planning on baking textures, but I get that for texture baking, VERY high poly models are necessary.

Kitteh: That's typical game models, isn't it? For models used for stills and other stuff, I've mostly seen 20k-100k poly models, but without baked normals.

Ghostscape: That's really helpful input. That was actually what I was planning to do, it seems more logical to just chamfer some edges where needed for close-ups instead of going the sub-d route and trying to remove unwanted geometry afterwards. Also, I see what you mean regarding the exaggerated edge softening, on the one hand, this allows for the capturing of the textures and the later use on the low-poly model, on the other hand, this usually gives the in-game models a "plastic" look associated with modern games. So I'll probably just add some chamfering to the edges where it's needed.

PredatorGST: Interesting thought, although I've never heard that this is done usually. I think this is moot mostly due to the fact that for film work, render time is less of an issue compared to game work, so realism is higher on the priority list.

Cheers!
useless