Time steps in cloth simulation

[SimonDM]

Hello,

I've been reading several papers on cloth simulation and particularly implicit versus explicit but I rarely find it mentioned exactly how much bigger the time step can be (practical maximum) or is (in practice) taken than the maximum allowable time step in explicit simulations (Courant Friedrichs Levy criterion).

[SimonDM]

The other thing I'd like to know is because I see mentioned that conjugate gradient has a time complexity of O(n3/2) fortwo dimensional problems as cloth, while as for explicit integration this is O(n), so all things equal this would make a system with 10.000 points still 10 times slower even if a 10 times bigger time step would be taken. Or is collision THAT big of a problem or stability THAT much of an advantage for implicit? By the way, I'm working on an implicit method with complexity O(n) (which I think will work for square pieces of cloth for starters), would that be some sort of holy grail for cloth simulation?