another question on efficiency

Hi all, another question on efficiency: which is better? (1) a model which is broken into several/many trianlge strips and triangle fans. Create a big array for the vertices, and for each strip or fan an array of indexes then call glDrawElements for each set of traingle strips/fans (2) the same model, but instead of being broken into strips and fans, just has a listing of traingles, and call glDrawElements once to render it. I basicly have in mind quake2 md2 models in mind, which are typically less that 1000 triangles anyways... Best Regards.