Does computational power affect the final quality of I-TASSER models?

Does a slower computer just take longer to generate the output or can it make the output lower quality due to a "maximum time limit" of some type for simulations?

I am asking because I have noticed a general trend (not confirmed of course) of lower C-score structures from the I-TASSER distribution outputs when compared to the online server.