This would still consume resources, and indeed, if limited resources were somehow improperly allocated in a catastrophic manner, it might tear apart the fabric of the universe and threaten to crash both simulations.
But think of some of the implications in this. First: when that simulation attempted to discover if it was wrapped in a simulation, it probably would get a false negative, because it'd be running so close to the hardware. Second: WE wouldn't be able to tell if WE were slaved by another simulation. That is, whether we had a parallel universe as a neighbor that was pulling our strings. Third: If we find ourselves creating simulations that are alarmingly convincing, and seem to prove that WE are a simulation, we would want to be very careful when we start playing with one, for fear that we might have unwittingly stumbled upon a curious vulnerability in this realm that allows for an injection attack, since it might crash the system (although, if this is all just a big video game, what is there to honestly fear...).
See also: Bobby Tables: http://xkcd.com/327/
I mentioned the Halting Problem as an example of Undecidability. We know that there are problems that cannot be solved by a computer, regardless of the resources such computer may have. The argument that we are a simulation would only apply if all observable phenomena in the universe are computable themselves (which is far from trivial to answer, but if I had to guess I's say those aren't). Of course, we could argue that whoever ran this simulation would have provided it with a simplified reality, including an underpowered form of computing... but that sounds rather suspicious to the skeptic in me.