Thanks for the answers guys, If i base it by the number of tris, thats just the number of vertexes/3 so it would'nt make too much difference really lol + the way OpenGl works, when i render each vertex is placed seperately, so i would'nt be able to load it, because i ionly haev the Tris saved, if you get what i mean.
The Engine so far does have a culling system built in so it's quite efficient in that respect. The problem with me testing is,i have a very good system, its well above a normal pc (not to show off) and it really can take pounding. I had 1004 objects which all had 65000 Vertexs and it didnt sem to be slow at all! i dont have a method to test the FPS at the minute so ill get back to you on that 1 but it was definately playable with that if it were a game that is.
So far i think the best method may actually be to make seperate objects of size (small med , large etc...) but hopefilly i will find a way to dynamically recreate and destroy the array which holds each vertex, which means i may be able to hold an 'unlimited' amount of vertex in each model... and by that i mean no limited by the Engine.
-BEN