Basically any single DOF and/or texture needs to have a draw call sent to the CPU to prep it for rendering.
So if you have a DOF with one texture object inside it, that means a draw call.
Or if you have one DOF with say 5 textures in it, then that is 5 draw calls.
So it's quite common perhaps that 5 fence objects each with 2 textures per object, will need 10 draw calls in the end.
In that case it might be better to have 2 fence objects, one big one for all the first texture, and one big one for all the second texture, so two draw calls.
Really a draw call should contain as much data as possible without causing costs elsewhere.
So lets say you have a 1/4m drag strip where all the fences are visible all the time, probably best to have one big object.
But on a long track where you can't see some fences, then split them up, but then turn those fence objects off when they can't be seen.
If you chop up fences like this for example, you'll always see two, the one you are 'in' and the one coming up ahead... so no point splitting if you are on a small track where you only need to split once any way... may as well have no splits than one split and still always see all the fences most of the time!
Just think of what you can see and what can be sent to the CPU all as one job rather than many smaller ones.
Obviously there are lots of trade-offs and balancing to do, but generally if you try at all to optimise then the FPS should be good.
If you work really hard at it then the FPS should be increasingly good.
Console game optimisations go really far because on a fixed platform those smaller and smaller optimisations will be known to be a benefit to everyone.
On a PC that is less the case, and differences in CPU/GPU may swizzle the balancing around a bit, but it still holds true that optimising is key to good performance.
Like I said, on that Roggel track that was released by Cruden a few years back, the FPS on my machine were like 50fps or something. I just batch optimised the trees alone and my FPS went up to about 65fps, or a 30% gain perhaps.
Add in LOD, and batching all those other buildings etc and no doubt you'd get even more FPS.
I think with an optimal rendering set-up for a mid-range PC that track would have run at 100fps on my machine. It just takes time to do!
Dave