Hi Tim,
> I understand that many people are extremely productive at hard surface modeling using subds, <...>
It can vary a lot depending on the particular qualities of the model - "hard surface" is a rather generic term...
If the objects being created have the overall form of something like one connected skin, then the sub-d modeling can progress in a similar fashion as it's done for organic models and it can go well. Things like space armor and vehicles can fit into this kind of stuff. These kinds of things can be in the sort of "gray area" that I was mentioning previously where it's not always 100% clear whether you'd want NURBS or sub-d for them and when that happens you may actually want to err more on the sub-d side especially if you're still getting up to speed with NURBS modeling strategies.
If the objects being created are more like engine parts that have a lot of holes in them, then suddenly the sub-d productivity level drops off a very steep cliff even for expert users. There is just fundamentally a tremendous amount of manual topology arranging work involved to try and accommodate precise holes when using sub-d, and it also gets easy to have a lot of imperfections like little bumps or ripples - those are often unwanted side effects when you're trying to force a sub-d topology into place that has to deal with a lot of constraints.
Check out this thread on the modo forum for a good example of this specific type of situation:
http://forums.luxology.com/topic.aspx?f=32&t=67986
There you'll see an example that's really clear, a whole lot of topology planning and just overall work needs to be done to do that in sub-d but meanwhile in MoI the same model is done more precisely in literally 20 seconds of work, see the youtube MoI screencap video that's linked to in the thread there.
So a lot of this depends on the particular qualities of the model at hand. When you get accustomed to using MoI you'll be able to recognize the kinds of things that will come together really quickly in it.
re:
> Therefore, a typical workflow for someone creating a game asset for Unity using subd modeling
> would be: create high res subd model, create low res version of same model,
So yeah it sounds like you've already got this figured out, but since you are not talking about retopo to produce quad topology for making sub-ds and are more focused on just making low poly, it is certainly worth a shot to just generate both a high poly and a low poly output from your same MoI model and see if that does the job for you at least for some particular cases.
NURBS models have a similarity to sub-d models in the sense that they define some "ideal smooth surface" and you can generate different approximations of that at export time, either high density or lower density output from the same base model.
- Michael
