I think you're asking, if we took all the matter in the Universe and converted it into a giant computer, would we be able to fully simulate our own Universe?

I think the answer is that if our Universe is currently, say, nested N levels deep in a simulation hierarchy, the Universe we currently observe may appear much more vast than it actually is, but since the speed of information is limited, we could never prove this. So while the number of estimated atoms in the observable Universe is something like 4×10^79, the actual number of simulated particles may be far, far less. Certainly, the information that comes to us via the light from distant galaxies is incredibly coarse grained and time-lagged, and if we're in a simulation it may well be that no finer grained representation actually exists (assuming there are no observers in those galaxies).

I don't actually know if I believe the simulation argument, and all of this is currently (perhaps forever) non-falsifiable. By the time we're able to even send an observer to some distant galaxy ~5 billion light years away, the Universe would have expanded even more, making the galaxy even further away. But the limitations imposed on the speed of information via light cones seems to make it very computationally feasible to simulate a Universe that appears much larger and detailed than it actually is. So in other words, the inhabitants of Universe at level N of the simulation hierarchy should be able to simulate a lower-entropy coarse approximation of their Universe at level N + 1, that would be indistinguishable from Universe N to the simulated inhabitants of N + 1, provided that sufficient restrictions on the speed of information and the number of observers is enforced.

But intuitively it seems there should be a limitation here, in that as N increases the level of nested computational resources available would need to decrease, until you hit a low entropy "leaf Universe". But that doesn't really affect the probabilistic angle of the simulation argument.