How Much RAM Does a Project Zomboid Server Need?

On our hardware, a freshly generated vanilla world with nobody connected settled at roughly 2 GB. On a 2 GB plan that is not a comfortable floor, it is the ceiling — it left effectively nothing spare with no players and no mods, which is why we no longer sell a 2 GB tier. 4 GB is the smallest plan that actually clears that floor with real headroom, so that is now the Starter tier. Memory use grows further as players explore and the server keeps more of the map loaded, so the right question is not what an empty server uses but how much room it has left once your group has been playing for a few weeks — for a larger or long-running world, that is what the Standard and Pro tiers buy you.

Because an empty world simply does not request the memory, so what you are seeing is a property of the test rather than of the tier. The differences between tiers at idle were smaller than the variation we saw between two identical runs of the same configuration, which is another way of saying there was no real difference to measure. Once a world has actually been played in, the same server rose to roughly 3 GB. The idle figures show each tier starts cleanly; the played-in figure shows where usage actually goes.

Not directly as game memory. Project Zomboid runs on Java, and the server allocates its Java heap as a proportion of your plan's memory — about 60% of the tier. So a 4 GB plan gives the game roughly 2.4 GB of heap, with the remainder covering the Java runtime itself, the operating system's file cache for map data, and the safety margin that stops a spike from killing the server. A bigger plan raises that ceiling even when a quiet server never reaches it.

The server process is stopped by the host and restarts, which players experience as a sudden disconnect. It is worth being clear that we did not see this at any tier we measured: across every run on this page the host recorded zero out-of-memory events. That is the specific thing the tiers are sized to avoid. One honest limit on that result: it counts servers that were forcibly stopped, so it cannot speak to a rarer situation where a server keeps running but stops making progress while the machine it sits on is under memory pressure. Nothing is stopped in that case, so nothing is counted. We track that separately rather than treating a clean count here as proof it cannot occur.

No, and we are not going to guess. Every figure here is from a vanilla server, and mods change memory use significantly — a large map mod is a different workload from a small quality-of-life mod, and the difference between them is far too large to estimate honestly. We would rather publish the vanilla numbers we actually measured than a modded number we did not. If you are planning a heavily modded server, start with more headroom than the vanilla figures suggest. Worth knowing before you plan a mod list: Build 42 requires mods updated for Build 42, and a good number of popular mods have not been updated yet — a Build 41 mod will not load. If your server depends on older mods, contact us and we will talk through the options.

Each tier ran on the same server on our own hardware, on Project Zomboid build 42.20, with a freshly generated world each time and only the memory limit changed between runs. Peak usage was read from the Linux kernel's own accounting for the container rather than from inside the game, so it reflects everything the server actually used, and each run was left to settle after startup before the reading was taken.