Why you should double edge loops in sub-d
interpolator
I remember reading the sub-d thread a while back, and racer445 said something about doubling edge loops and how it makes sub-d models looks more professional, well, it's not just that, it's also practical. People asked and he didn't really show why you should, so I thought I'd share my findings. I don't know if this is common knowledge but I'll show anyway,
This is why I think it's helpful, it's not nessisary but it can help improve veiwport performance with very complex meshes in my experience. Not everything.
SO, with that said, here is what it looks like when you have your wires(crap wires) with single edge loops, unsmoothed
and next is single edge loop but with nurms set to 1 iteration, or turbosmooth iteration 1:
as you can see there are issues, and overall it looks pretty bad, you naturally you'd set it to a value or 2 or 3, to get this look:
See but then the issue arises that this is now using a larger polycount, that will decrease veiwport speed with complex models, having every or almost every object at a smoothing or 2-3, thats a lot of polys.
Now lets look at what happens when we double the edge loops, here are the wires for the unsmoothed, you can see that they are doubled, one extra edge behind each support edge(except of the middle parts):
Now here comes the awesome part, and the reason having the edge loops doubled is important. This is the same mesh with a numbs iteration of 1:
As you can see, it's the EXACT same result as that of the single edge loop mesh at 3 iterations, but being able to get the same results at one will me a lower veiwport poly count, thuse allowing more to be on screen and still look great. I have a great system, (3930k, 24 gigs ram, 7970 3 gb ect) and my veiwport would lag because I would have to set it to 2-3, well with most objects you don't have to, and not only will it look better, but save you time, and allow you to be more efficient, and I'm sure it would make render-times faster as well since it's a lower polycount.
Well, that's about it,I hope this was of use to someone, if you have any questions let me know, or if you wanna critique what I said, this is just what' I've noticed. Sorry I didn't really present this with pretty stuff, I just kinda felt like making this, if enough people find this helpful I will attach better images.
This is why I think it's helpful, it's not nessisary but it can help improve veiwport performance with very complex meshes in my experience. Not everything.
SO, with that said, here is what it looks like when you have your wires(crap wires) with single edge loops, unsmoothed
and next is single edge loop but with nurms set to 1 iteration, or turbosmooth iteration 1:
as you can see there are issues, and overall it looks pretty bad, you naturally you'd set it to a value or 2 or 3, to get this look:
See but then the issue arises that this is now using a larger polycount, that will decrease veiwport speed with complex models, having every or almost every object at a smoothing or 2-3, thats a lot of polys.
Now lets look at what happens when we double the edge loops, here are the wires for the unsmoothed, you can see that they are doubled, one extra edge behind each support edge(except of the middle parts):
Now here comes the awesome part, and the reason having the edge loops doubled is important. This is the same mesh with a numbs iteration of 1:
As you can see, it's the EXACT same result as that of the single edge loop mesh at 3 iterations, but being able to get the same results at one will me a lower veiwport poly count, thuse allowing more to be on screen and still look great. I have a great system, (3930k, 24 gigs ram, 7970 3 gb ect) and my veiwport would lag because I would have to set it to 2-3, well with most objects you don't have to, and not only will it look better, but save you time, and allow you to be more efficient, and I'm sure it would make render-times faster as well since it's a lower polycount.
Well, that's about it,I hope this was of use to someone, if you have any questions let me know, or if you wanna critique what I said, this is just what' I've noticed. Sorry I didn't really present this with pretty stuff, I just kinda felt like making this, if enough people find this helpful I will attach better images.
Replies
I highlighted the same edge loop on each of the 4 meshes. The group on the left is with only one support loop and as you can see the edge gets moved pretty far away when you sub divide it. The group on the right is with an extra support loop and as you can see the edge barely moves. However if you look at the vertical support loops they do not move that much and you would probably be ok without any more vertical edge loops.
So it really depends on how you want to control how sharp you want your edges really. Simply saying adding double support loops to everything isn't really understanding what Sub-D is actually doing.
I wasn't saying it to everything either, and I didn't add them to everything in my example.
It wouldn't really matter if you were say modeling a gun, but I'm modeling something that 1/4 of the unsmoothed mesh is 1 million polys at the moment (one of the walkers legs) and little things like this really make a difference for me in terms of performance. Even having everything at 2 is a big difference and is considerably lower. That is just one of the legs and with 4 legs and the base body I can barely navigate the veiwport without extreme lag. Yes, I don't model with everything on screen but I think that in very complex and high poly meshes it's helpful. So that in effect is why I think you should. For example, http://www.polycount.com/forum/showthread.php?t=103629&highlight=lonewolf
I agree with earthquake here.
Trying to establish rules for highpoly modeling is a waste of time. If the end result looks good then nobody cares. You're not putting it in game. It's just source for a render.
I personally just have a different render setting for my meshes so I'm not actually seeing the final result for all objects in the viewport. Then I don't have to worry about using too many polygons in the hipoly.
But I don't do it on a compulsory basis. After all there are times when you want long smoothed edges instead of a very small smoth.
And like I've said numerous times, not trying to create rules, trying to answer a question asked a few months back in the subd thread.