computing power required to run a universe simulator

Look into the computing power it would take to simulate a universe to the detail level that this one is simulated at with the power saving and without Short answer: the two figures differ by roughly 211 orders of magnitude, and that gap is the actual result — the cost isn't a number, it's an interval whose position is set entirely by assumptions nothing currently constrains. Using "detail level we observe" rather than "detail level it's simulated at," to keep the programme's usual separation intact. Without savings — classical Planck-resolution lattice Observable volume ~3.6×10⁸⁰ m³ ÷ Planck volume 4.2×10⁻¹⁰⁵ m³ ≈ 10¹⁸⁵ cells Age 4.4×10¹⁷ s ÷ Planck time ≈ 10⁶¹ ticks Total ≈ 10²⁴⁶ updates, or ~10²²⁸ ops/s to run in real time Lloyd's bound says our universe has performed about 10¹²⁰ operations since the big bang. So a Planck-resolution classical run of this universe exceeds the total computational capacity of a universe like this one by ~10¹²⁶. You need a substrate categorically unlike the thing being simulated. arxiv If you demand the full quantum state classically it's worse in a way that stops being a number: dimension 2^(10⁸⁰). Not expressible in Planck volumes. That branch dies immediately unless the substrate is itself quantum — in which case ~10⁸⁰–10⁹⁰ qubits suffices and there's no crisis at all. With savings — stacked, largest first Locality (finite c): neighbour updates are O(N) per tick, not all-pairs O(N²). A fully-connected version goes as ½(V₃cT/l⁴)² ≈ 10⁴⁹². So c already buys ~10²⁴⁶ for free. arxiv Holographic bound: ~10¹²³ bits, not 10¹⁸⁵ cells — area not volume. ~10⁶². Coarse-graining to probed resolution: Beane–Davoudi–Savage bound the inverse lattice spacing at b⁻¹ ≳ 10¹¹ GeV from the high-energy cosmic-ray cutoff — i.e. ~10⁻²⁷ m, eight orders above Planck. A lattice sitting right at that limit is ~10³² cheaper in 4-volume. arXiv Vacuum sparsity: matter occupies ~10⁻³⁰ of volume at atomic density. Observer-relative rendering: compute only what enters records. ~8×10⁹ humans × ~10⁷ bits/s ≈ 10¹⁷–10²⁰ ops/s — under a gram of Lloyd-optimal matter (a 1 kg ultimate laptop runs 10⁵¹ ops/s). The correction worth flagging The standard "collapse-on-observation is lazy evaluation" move has the complexity sign backwards. On a classical substrate, an unmeasured superposition costs 2ⁿ amplitudes; a definite outcome costs n bits. Nature carries the expensive object right up to interaction and then discards it. That's eager evaluation with garbage collection, not lazy evaluation. Decoherence is a deallocator, not a renderer. The features that do point the right way are the boring ones: finite c, the holographic bound, and — strongest — renormalisation. Effective field theory means short-distance physics genuinely decouples, so you can compute low-energy behaviour without Planck-scale input. Physics is natively level-of-detail structured. That's a much better "power saving" argument than collapse, and it's already standard physics. For RCC This is Task E's verdict at cosmological scale: 10²²⁸ vs 10¹⁷ ops/s isn't two estimates of one quantity, it's proof that cost isn't a function of the world being simulated. It's a function of the algorithm and the fidelity target, neither of which is observationally fixed. Any paper quoting "the compute needed to simulate the universe" as a single figure has smuggled in an algorithm. The one tier that's actually discriminating is the lattice one — UHECR arrival directions would show rotational symmetry breaking, but only if the spacing sits near 10⁻¹² fm; smaller and there's nothing to see. Worth checking whether Auger/TA anisotropy data since 2014 has moved that bound, since it's the only place this whole analysis touches an observable.