ArmaDB
Sign in

Arma Reforger hosting calculator

RAM, CPU and disk for an Arma Reforger server, read off the servers already running with a similar modpack and player count. Measured, not modelled.

Modpack

or

What to rent

12 GB RAM 3.5 GHz+ core 13.0 GB disk 500 Mbit/s uplink

For 48 players at peak on 64 slots, vanilla.

RAM

12 GB

The numbers ask for 8 GB. 12 GB is what you rent for 64 slots: room for headless clients, AI and a bad night.

CPU

60 server FPS

The simulation runs on one thread: 3.5 GHz minimum, 4.5+ to hold 60 when full. Core count past that buys nothing.

Disk

13.0 GB

Mod sizes from the Workshop; the install size is an assumption.

Bandwidth

500 Mbit/s

14 times the peak. Mod downloads go through the Workshop, not your server.

Measured on 2,315 server-hours from 111 servers online in the last 7 days, with no mods and 36 to 60 players in the hour.

Servers like this, right now

memory in use as reported by the server process
How this is calculated, and what is guessed
RAM: measured, then padded, then floored
The 90th percentile of what comparable servers report, hour by hour. That is the game process, not the machine, so we add a quarter for the peaks an hourly average hides, 1 GB for the OS and up to 4 GB of file cache. Startup is a floor, not an addition: a server appears in the browser only once it has loaded, so we cannot see that phase (checked on 13,481 sessions: after a server appears, the first twenty minutes match its steady state). Hosting guides put it near 3 GB, and loading never happens during a match. Last comes a floor by slot count - 8 GB up to 32 slots, 12 up to 64, 32 above - which is deliberately higher than anything we measure: it is what people actually rent for a server that size, with room for headless clients and for whatever else ends up on the machine. Whichever is larger wins, and the card says which one decided.
CPU: server frame rate measured, clock guessed
The frame rate here is the server's simulation rate, not what a player sees: a server at 30 with a client at 144 is a stuttering game. The hours are real, with anything above 1,000 dropped as an empty server. No core count is given, on purpose: the simulation is one thread, and the measurements say load barely moves with player count - the median holds at 60 up to 99 players and only falls to 54 between 100 and 128. Anyone selling you a core per N players is guessing. The GHz figure is a recommendation too, since we can see frame rates but never the processor.
Disk: mods measured, the rest assumed
Mod sizes are the Workshop's own, exact when you pick a real server. About 12 GB for the install and room for one update of the largest mod are assumptions.
Bandwidth: not measured at all
Traffic between a server and its players never passes through us. The per-player rate comes from what hosts publish (a 32-slot Conflict sits between 5 and 15 Mbit/s, a modded one with heavy AI passes 30), hence the range. What you order is a port, so we name the smallest on sale that covers the peak three times over. Mod downloads are not included: the client takes them from the Workshop.
What it cannot tell you
AI count and mission scripting move the CPU more than the mod list does. Headless clients are not counted, and each one is another instance of the game: two of them roughly double the memory. Nor is anything else sharing the box.