What is better and may optimize engine speed

a) using a lower polygon size or
b) using a lower bitmap size

Say I have an obect of some 750 vertices, which is around 45 kb (without picture).
A twofold precise model of 1500 vertices would have 120-200 kB which is not so much compared
with the difference of a bitmap

256*256 = 192 kB
512*512 = 768 kB

If I use high vertex resoultion model and 256*256 I would end up in a modelsize of max 390kb, whereas
low vertex resoltion and 512*512 would at least cost 800 kB.

So - what is the engine doing with that?

Memory-wise high poly models are more efficient than large textures (you need to multiply the texture size by 1.5 when using mipmapping BTW).
However, vertices need to be transformed prior to being displayed on screen. So depending on your CPU / graphics adapter, one or the other solution might be faster.

thanks for your answer, Marco Grubert (and to let a topic so relevant not disappear without any trace) Regards MB