It would be interesting to see someone with the background to understand the arguments involved in the paper give it a good review.
That said, I've never brought the simulation hypothesis on the simple grounds of compute resources. Part of the argument tends to be the idea of an infinite recursion of simulations, making the possible number of simulations infinite. This has one minor issue, where are all those simulations running? If the top level (call it U0 for Universe 0) is running a simulation (U1) and that simulation decides to run its own simulation (U2), where is U2 running? While the naive answer is U1, this cannot actually be true. U1 doesn't actually exist, everything it it doing is actually being run up in U0. Therefore, for U1 to think it's running U2, U0 needs to simulate U2 and pipe the results into U1. And this logic continues for every sub-simulation run. They must all be simulated by U0. And while U0 may have vast resources dedicated to their simulation, they do not have infinite resources and would have to limit the number of sub-simulation which could be run.