Here -- http://www.crydev.net/viewtopic.php?p=819231#p819231 -- is a post for bringing SketchUP models into the new crysis (game) engine. Even if you're not interested in the crysis game engine (although it has really cool potential for realistic walkthroughs) or even rendering engines in general it has some excellent advice about proper modeling technique in SketchUP. It's pretty well written and includes a bunch of clear images illustrating each concept.

The main things I would take away from it are:

How and why to pay attention to which direction faces are facing
why and when to use groups
how to apply textures and basic manipulation of these

Just to be clear, 3. General advices on modeling is the section I thought would be helpful, although the rest of the topic would be interesting/useful if you have any desire to learn how to prepare a model for most any rendering program.

I also like that sketchup does NOT have a rendering engine built in. As already mentioned, there are better, purpose-built engines already that integrate very well with SU. I doubt google would be able to improve on them and even if it did it would be at the expense of any other development. I would probably still use thea even if they did come up with a competitive rendering engine.

The the areas that I would like SU to improve are:

1. support for the ruby community. I am only an end-user so I can't begin to say what would be helpful, but the more accomodating SU is in this regard, the better.

2. Core modeling improvements. Better selection tools, more accuracy, subdivision, better ability to detect (and correct) enclosed areas (or areas that are MEANT to be enclosed)

3. NURBS modeling IF it could be implemented in-line with the pushpull methodology. Maybe it doesn't need to be NURBS (that's just what I'm familiar with) but I would like geometrically "pure" arcs/circles rather than the polygonal, faceted versions we currently have. The "on-surface" plugin seems largely able to do this, it would be nice to see it more integral, though.

This list makes me interested in looking into MOI3d again. Hm....

and I assume most people know this but...

if you scale by -1 then you mirror the object.

I usually use a texture for glass in Sketchup as a stand-in for a bump in rendering programs (Thea). No glass is perfect -- slight warping and imperfections. I usually use water textures scaled much larger (5 or 10 times larger than I would use it for actual water) and then a VERY slight bump setting, just to subtly bend and warp reflections.

Bad design for a keyboard, maybe, but why are keyboards so often overlooked? It's the piece of computer hardware we have the most direct connection to and everyone seems to go for the cheapest one ever. Every (arch) firm I've worked for have always just used the free Dell (or equivalent) keyboards while spending thousands every other year on new systems.
I got a nice keyboard at home but don't feel wealthy enough to buy one for work also -- although maybe I should considering the amount of time I'm here using it.
Maybe this should be a separate thread but I'd be curious what keyboards people here (high percentage of designers?) use.

I'd been toying with the idea of getting an ipod touch. I like the idea of having a phone and an ipod in one but don't see myself using any of the 3g (or whatever) functionality (I would rather get apps at home and never feel interested in checking the internet when I'm traveling) -- especially when it means paying for the data plan. I seem to belong to such a niche market, though, that there isn't an option to buy a "smart phone" with no data plan.
Anyways, the ipod touch looked as close as I could find to what I wanted but then I stumbled across archos' tablets.
It seems to be the android version of an ipod touch. No itunes.
Did you look into these before getting the ipod touch? is there some reason I'm not seeing that makes them inferior?

No reflection -- that must be a little vampire boy!

Nice render, though. When you said you wanted it to look 'aged' did you mean an old structure or a scene from the 30's?
My only criticism is that the sidewalk looks tiled... kind of. It's not an obviously repeating tile but it seems... repetitive.

Looks great, overall.

About the AMD (no longer ATI) issue: I've been using an HD6950 for several months and haven't noticed any issues with sketchup. I haven't run any tests to objectively say how fast it runs, but it's much quicker than my Nvidia 9800GT at work -- part of that is due to the processor, though (core 2 quad @ 2.66 ---> i5-2500k @ 3.7).
The point is, no problems with new AMD.

That does look great but I would also be curious about being able to export HDRIs. For me that would make the difference between an interesting program and one that would a part of my work-flow.

This must be the coolest use of java ever! Get ready to waste some time.

I've recently built a new system and haven't had much time to use sketchup on it until the last few days. I've been noticing a little stutter between orbiting commands. The display is very smooth but if I middle-click to orbit a couple of times in quick succession (say... less than a second between clicking) the second click won't register. I suppose it could be the graphics card (ATI) but it doesn't seem to be a visual problem. I just have to pause for a second before clicking again. At this point I haven't used it enough to notice if it happens with other types of commands.
Has anyone else encountered this? Does anyone have a suggestion / solution?

If you guys combined forces you would have the ultimate in military-geekdom (I mean that with absolute respect): the helo-mounted 88! I'm sure Jerry Bruckheimer would drool at the thought of it...

I don't have time to model an example but my first thought would be to copy the plane vertical to this thickness of the timbers desired and a second copy arbitrarily higher. Get the grid pattern on the higher plane and offset for required width (so that it's correct in 2d). Delete faces of this plane and extrude (I love the extrude line plugin for this) the remaining line-work through the 2 lower planes. intersect everything and delete garbage lines/faces. This should leave you with the desired, squared, gridshell.

You could also switch over to blender and use this tutorial.

When it's done just export it into sketchup.

I just built a system with an AMD 6950 and have had no problems at all. Turning on shadows/AA doesn't slow it down at all -- although I haven't tried any ridiculously large files. This is barely comparable, but I have a 9800 GT (nvidia) at work which works fine, but is very noticeably slower than the AMD card at home.

The main reason I got this card, though was for its 2gb memory for extra head-room with GPU rendering. That said -- and as satisfied as I am with it -- the GTX560 looks like a capable alternative (supports CUDA which AMD doesn't but only has 1gb).

The stand-out feature of sketchup, to me, is the inference-engine and it's snapping. I realize that this also puts limits on what (due to size) you can work on, but This is why it's still my modeling program of choice.
Several times I've thought to switch to blender or some other program, but I can't get over how easily something could have been done in SU that I've spent forever on the new program.
I do agree that the modularity of plugins are very nice. Still, at a certain point other tools can simply do something better than sketchup and it's easier to use those than fight sketchup to get it to do what you want.

I've just started trying to import models into crysis and thought playup would be the perfect plugin -- I still think it would be if I could get it to work.
I always get the wxruby missing error (this is using SU8 on win7 64-bit). I've tried manually installing wxruby but that doesn't seem to change anything.
I tried installing it on a a copy of SU7 and that worked once -- so I thought it was fixed -- but every time since it get's the same error as above.

I'm posting this not so much to complain (although if someone has a solution I'd love to hear it) but to mention a work-around that seems to work:

Model units set to decimal, meters
TIG's obj exporter after purging (haven't tried using groups or components so I'm not sure, yet, if those would be a problem)
Blender import of obj (using blender 2.56)
export as collada
converted to cgf (crysis format) with colladacgf

After that the model was imported into sandbox with no problems. One odd thing is that I used a jpg in SU (sized 512x512) and crysis (warhead) worked just fine with the jpg -- I thought it only could use dds?
Anyway, I've only just gotten this to work this morning and look forward to trying it more but I'm pretty excited about it.

This might be a little late of a response but maybe it will help someone or at least be a useful reference.

I just built a system around an i5-2500k and an AMD HD6950. I've played with it all weekend (although not exhaustively yet or even had a chance to benchmark anything) and have had absolutely no problems with sketchup or any other program. I'm not sure if it's the processor or the GPU but performance is (subjectively) very impressive.
From my point-of-view the ATI (now AMD) su errors are a thing of the past. If I do come across any problems after I have a chance to use it a bit more I'll let people know.

I think that the way DCs were implemented is a little more complicated that they should be. Not complicated as in difficult to understand, but rather tedious. To make a window frame that can be stretched, for example, you need to have a minimum of 8 objects with various interconnected parameters (4 corners plus 4 straight edges connecting them). When they were first released I had hoped it would have been implemented similar to (but with improvements) autocad's DCs -- to stretch something you draw a box around what should stretch (just as you would select it yourself) and add a grab tag. Much simpler.

I haven't seen almost any DCs other than VERY basic ones (color change, or on/off, open/closed type of behavior). It's a shame; I had been hoping to see a breakthrough of cool new components.

It IS still worth learning -- because it's very easy -- but I wouldn't expect it to change your work-flow drastically.

If someone has had a different experience with me and knows much betters methods for making (automating the production of) DCs please post.

I had a question that I'm hoping someone can answer. I've been starting to use the arbaro tree generator and am pretty happy with it, but when I bring it in to blender (.obj) I can't figure out how to texture the leaves. I was hoping they would be instanced (however blender would handle/call that) so that I could map 1 leaf and have the rest follow suit but either its not working or I'm doing it wrong. Then I thought I might be able to stack all the similar (exact same) leaf meshes in the uv editor and apply an image to the lot of them, but that doesn't seem to be working either.
Any ideas as to how to do this or is there a better way to texture a tree?
I've watched this tutorial on making trees -- and it seems like a decent alternative, although I'd still prefer to use arbaro -- but it doesn't address my initial question.