For the current Racer engine, it's shader/draw count that will hurt performance more than VRAM use per se.
So combining textures and meshes so the renderer can deal with them in one pass is more preferable to having lots of smaller bits.
As per shading in the texture and liveries, I'm not sure what the best approach is, but ideally unwrapping for liveries is good because it also lets you add ambient occlusion data and also possibly a mix map for other things if you like, such as dirt maps or whatever.
I think unwrapping and having an ambient occlusion pass adds a lot to cars, since those areas in the shadows will come alive a lot more.
But then the base colour map really is wasted a bit if it's a solid colour.
But then a solid colour is good for liveries, since you need RGB for that.
As an example of what I'd personally do for a road/race car with livery, is I'd do the full unwrap as usual.
But perhaps try keep that texture to just the bodywork/solid paintwork elements, so if you have a 'white' version, it can be multiplied in the car.shd for different colours, without making other parts tinted (ie, if there are other colour parts in that texture/shader it'll be tinting those parts), or if you have a coloured version it can have livery.
OK you could put other bits in that texture map too, say indicator lenses, whatever else, but it'd be in a different shader that didn't get colour adjusted, and there is no point having it all in one big texture to do that, unless you feel there is 'free' room in that texture that is otherwise wasted.
In the end what you're doing is all good work assuming you keep PSD files for all the textures, then if we figure out better approaches later we can just swizzle textures around and apply superior shading. The base textures/UV maps etc generally all tend to work fine.
Ie, in your tyre above it's shaded darker to the back under the car. Really you could do that via an AO map, so in full sun it doesn't have a dark band at the back with the wheel turned.
But in the shade it'd look darker towards the back. Subtle difference but with modern shaders these are the things we can easily do, and it'd just be a matter of tweaking the textures for a bit to achieve that result!
As per normal maps, are you using them even on the main bodywork? I think for those kinds of jobs these days it's vogue to run more polygon detail and then just run lower poly models on LOD further away.
But for interiors and the matte parts of car exteriors I think normal maps really help as there are loads more 'fine' details that normal maps can cover easily. Ie, stitching on steering wheels and seats, or the subtle texture of dashboard materials, and of course tyre treads
I'll try get a WIP up soon, and also finish the project haha.
So far I'm using a photo/camera matching program to generate cameras and then using Max to do the spline work.
I'll then model from the splines using a sub-d approach at level 1, then collapse, then cut in details (like many high poly modellers do for later sub-division again).
That seems to get nice smooth surfaces, nice quad flow, but keep a moderate polygon level. Then just doing little bevels on all the panel gaps at the end.
I think that tyre approach is good with the middle for the wheel. But for my project I want to have lots of wheel swaps so you can choose the ones you want, so tyre/wheel will be split out I think.
I'll probably still run with the usual way in the end just because it's faff-free and ultimately easier for other people to re-use on slightly different tyre widths/profiles etc, as there is some 'space' around both tread pattern and sidewalls for doing some re-working or extensions etc!
Cheers
Dave