I'm looking for a simulation of different water models (TIP3P, SPC/E, TIP4P, water models with flexible bounds) where the differences in simulation speeds on a common architecture are really seen. Everyone says TIP4P is computationally more costly than TIP3P which makes sense as TIP4P has more degrees of freedom but I want to quantify it. Ideally I'd like to have a comparison between a water model with flexible bounds and one with rigid bounds but I'd happily take other sources as well.
You may be able to find some bounds on performance such as O(n3) based on the number of water molecules etc, but honestly benchmarks not performed on your own hardware are always somewhat hard to use or accept. Things such as memory speed can also influence things quite significantly so benchmarks are hard to look at for quantifiable differences.
That being said, trends may apply, but you already know this since you know TIP4P is more computationally expensive than TIP3P. If you want to quantify it, you really need to run your own benchmark. Simply create a relevant water system and run it with all relevant methods for a couple of iterations and compare.
If you do benchmark it for your system though, please add an answer sharing your benchmark and system. Sometimes trends can shift based on software version or major hardware differences.