← Dev Log

The Tier Loop: reliability you build, not buy

Uptime Tiers I–IV aren't a menu choice - they emerge every sim-tick from the redundancy you wired in. Here's how the model works.

The single idea the whole simulation hangs on: you never select an Uptime Tier. The site’s tier emerges from the redundancy you actually built, and it is recomputed every sim-tick.

Two ladders, deliberately independent

People kept conflating two different things we both called “tier,” so we split the vocabulary for good:

  • Uptime Tier (I · II · III · IV) - reliability, driven by the power and cooling redundancy you build.
  • GPU Gen (Hawk H100 → GH300) - compute, driven by the earnings you make.

They never gate each other. A Tier I hall can run the newest silicon; a Tier IV hall can run the oldest. Treating them as one axis made early builds feel railroaded - separating them is what made the sandbox breathe.

Redundancy is N+1, and it costs you

To hit Tier II you wire N+1 on both power and cooling - one spare of each, ready to take over the instant a feed trips. That spare capacity is pure insurance: it earns nothing while everything is healthy. The tension is the game. Every dollar of redundancy is a dollar not spent on a revenue-earning GPU.

Why the desert fights you

Cooling load climbs with the heat. Under-cool the halls and GPUs throttle, faults start, and a single fault can cascade into the neighbours - dragging your emergent tier down right as your fattest contract demands its SLA. Keep crew on hand, dispatch nearest-first, and watch the night call-outs.

That cascade is the subject of the next entry.


← All entries↓ Play the demo