Arma Reforger Server Requirements: RAM, CPU and Player Slots

What Bohemia does and does not publish

Bohemia gives you the server software, the config keys and the startup parameters, but not a proper sizing sheet. That is why host advice ranges from 4 GB to 8 GB for roughly the same advertised player count. If you only need the install side, see our guide to setting up an Arma Reforger dedicated server; the table below is the bit that matters when choosing a plan.

The baseline: what the server uses before anyone plays

The first reason Arma Reforger server requirements look vague is simple: the idle server is not tiny. Before you think about slots, AI or mods, there is already a meaningful memory bill to pay.

Look at the baseline first, then look at what room is left on the plan.

Across published host observations, a freshly started stock server lands at about 2.4 GB to roughly 3 GB even with nobody connected. Bohemia’s own stats example goes further: with only 2 players and 104 AI, the server is already at about 3.3 GB. That is why a 4 GB plan can work only for very light setups: you may have roughly 1 GB left after a roughly 3 GB baseline, which disappears quickly once the scenario starts doing real work. By contrast, a 6 GB plan leaves about 3 GB headroom, and an 8 GB plan about 5 GB.

You can watch this with -logStats. Bohemia’s example line looks like this:

FPS: 60.0, frame time (avg: 16.7 ms, min: 9.3 ms, max: 23.7 ms), Mem: 3291106 kB, Player: 2, AI: 104, Veh: 0 (17) ...

The Mem figure is memory in kB as reported internally by the server. It is useful for trends, but it is still approximate rather than a perfect match for what the operating system reports. What matters is not the exact last digit; it is whether memory climbs close to your plan limit when the server is loaded in the way your group actually uses it.

That baseline is the main reason “8 GB” and “64 slots” are answering different questions. One is a memory ceiling. The other is only a join cap.

What really loads an Arma Reforger server: scenario and AI first

Two servers can have the same slot count and wildly different load. The big driver is the scenario’s simulation work: AI, persistence, building, logistics, spawned content and how far the server has to keep the world relevant.

This comparison is the key point: same slots, very different server stress.

Conflict is usually the expensive case because it creates sustained CPU work. Base building, supply lines and persistent AI logistics keep the simulation busy across the whole session, not just during one short firefight. Game Master often starts lighter because it begins without pre-placed AI, but it can still grow heavy later if the session fills up with spawned units and objects. A PvP mode with AI disabled is cheaper again, even at the same slot count, because the server is no longer spending frame time on AI thinking, movement and pathing.

The core keys behind that look like this:

{
  "operating": {
    "aiLimit": -1,
    "disableAI": false,
    "disableNavmeshStreaming": false
  },
  "gameProperties": {
    "serverMaxViewDistance": 1600,
    "networkViewDistance": 1500
  }
}

• aiLimit: the default is -1, which means no limit. That is flexible, but it also means the scenario can keep creating CPU work until the mission logic or the hardware becomes the brake. Lowering AI mainly relieves CPU on the main simulation loop. The trade-off is obvious: fewer bots and less of the AI-driven feel that makes Conflict what it is.
• disableAI: if you switch AI off entirely for a PvP mode, the server gets much cheaper to run. The trade-off is that every feature depending on AI disappears with it.
• disableNavmeshStreaming: this loads the full navmesh into memory. It can give slightly better AI performance because the server is not streaming navmesh chunks on demand, but it costs more RAM, potentially up to hundreds of MB depending on terrain. Use it only when RAM has room and AI behaviour is worth the extra footprint.
• serverMaxViewDistance and networkViewDistance: higher values mean the server keeps more of the world relevant further out and replicates entities over a larger network range. That mainly costs CPU and bandwidth, not just memory. The trade-off is better long-range awareness for players versus more work every frame. Bohemia’s example config uses serverMaxViewDistance 2500 and networkViewDistance 1000, which is a good reminder that these are separate levers.

How do you tell whether a change helped? If you reduce AI or rein in view distances, the sign to watch is better server FPS and lower frame time under the same kind of fight, not a dramatic fall in idle memory.

CPU: frame budget first, core count second

The server does not care about your advertised slot number nearly as much as it cares about finishing each simulation frame on time. At -maxFPS 60, the budget is about 16.7 ms per server frame. At 120, it is about 8.3 ms. If the work for that frame takes longer, server FPS drops below the cap and players feel it as sluggish updates, delayed reactions and a generally soft, heavy server.

This is why single-thread speed matters so much in Arma Reforger server requirements. Parts of the workload can be spread across worker threads, and Bohemia does expose short-job workers with -jobsysShortWorkerCount, but the whole frame is still gated by the work that must complete inside that update loop. Extra cores help with supporting jobs, networking and keeping the machine from feeling boxed in, but they do not magically remove the main bottleneck.

That lines up with the published host view that the server is bound by single-thread speed and, in practice, does not use more than about six threads well enough to justify chasing ever-higher core counts for one huge shard. The practical lesson is simple: four fast, stable cores are usually worth more than a pile of slower shared vCPUs.

A shared or oversold vCPU hurts because the server is not only short on raw speed; it is short on guaranteed access to that speed. Even if the average looks fine, your frame time spikes when the host scheduler is busy elsewhere. The clue is inconsistent frame times at similar load, especially if memory usage looks healthy. Spare RAM does not fix that problem, because the issue is not capacity; it is late CPU time.

Mods and disk: memory pressure is only half the story

Before mods, the dedicated server is roughly a 12 GB download on disk. After mods, saves and logs, that figure stops being the interesting part. Mods live in the server’s profile area by default, they must be delivered to clients, and large terrain packs can push both storage use and runtime memory noticeably upward.

Terrains are the usual heavy hitters. They cost disk space first, then memory once the server has loaded them. Host-published storage presets in the 20 GB to 50 GB range make sense for that reason alone, and a multi-terrain mod list can need tens of GB of storage on its own.

The safest admin habit is to add mods one at a time. That works because it turns a fuzzy “the server feels heavier now” problem into a clear before-and-after change. Watch three things after each addition: idle Mem, start-up time, and peak memory during a real session. If you are running in a container and large mods hit write errors while downloading, -addonTempDir exists for that edge case.

Published host advice side by side

The host advice looks inconsistent because the assumptions underneath it are different. One host sizes by simulation first and treats slots as secondary; another publishes broad public-server bands; another points you at the idle baseline and tells you to watch memory in a real session.

The useful reading of that matrix is not “one of them must be wrong”. It is that 4 to 6 GB only makes sense for genuinely small, light servers; 8 GB is where about 32-player Conflict starts to look sensible; and the 12 GB to 16 GB range is where larger, modded or AI-heavier Conflict servers stop feeling squeezed. The reason the same advertised slot count can land in different rows is that Conflict, Game Master and AI-disabled PvP are not remotely the same workload.

Player slots explained: cap, queue and pricing

maxPlayers is just the upper gate. Officially it ranges from 1 to 128 and defaults to 64, while Bohemia’s own example config uses 32. That should tell you enough already: the right slot count is the one your CPU and scenario can sustain, not the default number in the file.

• Join queue: operating.joinQueue.maxSize ranges from 0 to 50 and defaults to 0, so it is off unless you enable it.
• Slot reservation: slotReservationTimeout ranges from 5 to 300 seconds and defaults to 60. This keeps a slot for someone who disconnected.
• Admin priority queue: since 1.3.0, up to 20 listed admins can get priority in the join queue.

Those settings matter because a full server does not always mean “buy more slots”. Sometimes it means “stop losing reconnects”, or “turn on a queue”, or “do not let a reservation window hold places longer than the group actually needs”.

Pricing models make this murkier. A slot-priced plan is selling access limits. A RAM-priced plan is selling capacity. Arma Reforger leans towards the second model because the simulation load does not scale cleanly with headcount. If you want the pricing logic behind that, read why slots and RAM are priced differently.

Conversion used: £1 = $1.33

Our starting-point sizing table

These are starting points, not official requirements. They are built from the published host floors and bands above, plus the known idle baseline. The reason 4 GB is missing is simple: it can work, but only for very light setups and with little room for mistakes.

If your server is mostly PvP with AI disabled, you can often start lower than the same slot count in Conflict. If it is Conflict with uncapped AI and ambitious view settings, assume the opposite.

Self-host or rent

Self-hosting can make sense for a private group, but only if you size the machine for two jobs at once. The Steam client’s 8 GB minimum and 16 GB recommended RAM figures are for playing the game, not for running the dedicated server as well. If the same PC is playing and hosting, add the server’s own roughly 2.4 to 3 GB baseline before a real session even starts, then remember that both the game client and the server are fighting for the same CPU time and storage.

Listen servers are started in-game, are PC only, and are best thought of as convenience hosting. A proper dedicated server runs separately on Windows or Linux and wants the official ports opened: 2001 UDP for the game, 17777 UDP for A2S query and 19999 RCON. If you are weighing the trade-offs, start with our guide to self-hosting a game server, then compare your spare machine against general game server hardware requirements.

Check it under load before you call it finished

One clean boot tells you very little. Run a proper session with -logStats 10000 so the server writes a stats line every 10 seconds, then read the session as a whole: idle, first joins, combat peak, autosaves and the late-session state. For deeper tuning detail, use our Arma Reforger performance and lag guide.

Tune it or upgrade?

Arma Reforger server FAQ

How much RAM does an Arma Reforger server need?

There is no official Bohemia figure, but the published baseline is the important clue: a stock server already uses about 2.4 to 3 GB before a real session gets busy. That makes 6 GB a sensible floor for a small vanilla group, 8 GB a stronger start for around 32-player Conflict, and 12 GB to 16 GB the safer range for larger, modded or AI-heavy setups.

Is 4 GB enough for an Arma Reforger server?

Only for very light cases. Once you start from a roughly 3 GB baseline, there is not much room left for scenario growth, mods, terrain overhead or long sessions. A 4 GB plan makes far more sense for a small PvP or light Game Master setup than for Conflict.

How many players can an Arma Reforger server hold?

Officially, maxPlayers can be set from 1 to 128, with a default of 64. In practice, the sustainable number depends far more on scenario design, AI and CPU speed than on the raw slot setting. A 64-player PvP server with AI disabled can be much easier to run than a smaller Conflict server.

Does an Arma Reforger server need a fast CPU or more cores?

Fast single-thread performance matters most because the server has to finish its simulation frame on time. Extra cores still help with worker jobs and general breathing room, but after a point they are secondary to per-core speed and stable CPU allocation. If server FPS falls while memory is fine, more RAM will not fix it.

Do more player slots mean I need more RAM?

Not directly. More players can increase load, but slots are only a cap on joins, while the heaviest costs usually come from AI, scenario logic, view distances and mods. That is why an “8 GB / 40-slot” plan can be a better fit than a “64-slot” plan that tells you little about the hardware behind it.

More Arma Reforger server help

Set up your server, fix lag, or compare hosts that fit your scenario.