Minecraft View Distance vs Server RAM: We Measured It (Paper 1.21)

What view distance actually loads

View distance is a radius in chunks (a chunk is a 16 × 16 block column of the world). The server keeps a square of chunks around each player, plus, on Paper, a two-chunk border outside it. That's why the count grows with the square of the setting: going from 8 to 16 takes one player from 441 chunks to 1,369, about three times as many.

Our runs matched that rule exactly for every single-player test and every spread-out group up to view distance 10. At 12 and 16, spread groups held 0.2–5% more than the rule predicts, most likely chunks near spawn that hadn't unloaded yet.

Simulation distance is the separate radius where the world actually ticks (mobs move, crops grow). We held it at 4 for every run, so this test isolates the cost of holding chunks.

The results

A straight-line fit through all 81 runs gives this formula. Feed it the chunk count from the rule above (not the measured count, which runs a little high for spread groups at 12 and 16) and it predicts every run to within 30 MB, usually within about 10:

live memory ≈ 237 MB + 0.1016 MB × chunks + 7.3 MB × separate player groups

(R² 0.999, where 1 is a perfect fit.) The 237 MB is the server's starting cost, close to the 228 MB the empty server used. The 7.3 MB is a small fixed cost per player area: the player and the entities around them.

Taking in the whole view at distance 16: about 140 MB of server memory for one player.  via GIPHY

Why grouped players are almost free

Ten players spread across the map at view distance 16 held 14,401 chunks and 1.7 GB. The same ten standing within a few blocks of each other held 1,444 chunks and 379 MB, about what one player alone used (378 MB). Chunks are loaded once per area, not once per player.

That changes how to size a server. A small group that builds one base together needs far less RAM than the same group exploring in different directions. If your friends scatter, size for each of them; if they stay together, size for roughly one.

Everyone in one room: a group standing together costs the server about as much as one player.  via GIPHY

What it means for your server settings

• Small group, one base

  View distance 10–12

  Players who stay together share their chunks, so a higher view distance costs about what it costs one player.

  Ten players together used 317 MB at 10 and 334 MB at 12.

• Group that explores separately

  View distance 8–10

  Every player who wanders off holds a full area of their own, so each step above 10 is paid once per player.

  Ten spread players used 942 MB at 10 and 1,723 MB at 16.

• Public or large server

  View distance 6–8

  Most players are apart most of the time, so memory grows with every one of them, and view distance is the setting that scales it.

  By the formula, each spread player costs about 71 MB at 10 and 52 MB at 8. We measured up to ten players.

• View distance 10 (the default) costs about 64 MB per spread-out player in chunks. Dropping to 8 saves about 19 MB per player; raising to 12 adds about 22 MB, and 16 adds about 76 MB.
• On a server where players travel apart, view distance is the biggest RAM lever you have after player count.
• If tick lag is your problem rather than memory, look at simulation distance and entities first; view distance barely moved MSPT here.

To turn this into a RAM figure for your setup, the game server sizing calculator now uses this measurement for its View Distance slider.

What we didn't measure

• One server build. Paper 1.21.11 build 132. Vanilla, Fabric and Forge weren't measured, so don't assume these numbers carry over to them; mods can store extra data in every chunk, which we didn't test.
• A pre-generated world. Generating new terrain costs extra CPU and short-lived memory that isn't in these numbers.
• Idle players. Bots stood still in creative mode at noon. Exploring players load and generate chunks continuously, and survival at night has more mobs.
• Live memory, not allocation. Between clean-ups Java let memory use rise a median 1.5 GB above the live memory we report (range 56 MB to 3.3 GB in an 8 GB heap). Allocate headroom.
• Simulation distance 4. At the default of 10 the server ticks more chunks, so expect more mobs and higher MSPT than here. We didn't measure that setting, so we can't say by how much.
• One machine. Server and bots ran on the same Apple M3 Max, so the bots took CPU from the server: treat our MSPT figures as an upper bound for that hardware.

How we tested

We ran 81 tests on 22 October 2026, each one a fresh server start, and measured live memory, loaded chunks and tick time after the players had settled. The raw results are one row per run: download the data (CSV, 81 rows). You're free to reuse it with credit to DevRunners (CC BY 4.0).