How much bandwidth does a game server use?
Rough per-player bandwidth figures by game type, what actually drives the numbers up, and why upload matters far more than download on a server.
6 min read · Updated Oct 21, 2026 · By GuavaHost Team
Two numbers people confuse
Bandwidth is how much data can move per second, measured in megabits per second (Mbps). Data transfer is how much moves in total over a month, measured in gigabytes or terabytes. A host might limit either, and they are different limits: you can be fine on one and run out of the other.
There is a third number that people blame for lag and that bandwidth has nothing to do with: latency, your ping. A fat connection with a long route still feels sluggish, and a thin connection next door feels instant. Lag versus ping pulls those apart, and if the symptom is delay rather than running out of capacity, start there.
Rough figures per player
These are order-of-magnitude figures for a server's upload, which is the direction that matters. Treat them as ballpark: the real number swings a lot with settings and with what players are doing.
| Game type | Example games | Upload per player |
|---|---|---|
| Blocky sandbox, 20 ticks | Minecraft Java and Bedrock | roughly 30 to 150 Kbps idle, bursting to several Mbps while a player explores new terrain |
| Survival sandbox with lots of entities | Rust, ARK, Palworld, Enshrouded | roughly 100 to 400 Kbps sustained |
| Smaller co-op survival | Valheim, Project Zomboid, Don't Starve Together | roughly 50 to 250 Kbps |
| 2D multiplayer | Terraria, tModLoader | roughly 20 to 80 Kbps |
| Competitive shooter, high tick | Counter-Strike 2 | roughly 200 to 500 Kbps per player, steady |
| Voice chat, in game | Simple Voice Chat and similar | roughly 30 to 60 Kbps per talking player, both directions |
Multiply by your player count for a rough sustained figure. Twenty players on a Rust server is maybe 2 to 8 Mbps of upload. Forty players on a Minecraft survival server sits comfortably in single-digit Mbps most of the time. These are small numbers by data centre standards, which is the point of this guide.
Monthly transfer follows from that. A server averaging 4 Mbps of upload, running all month, moves somewhere around 1.3 TB. Sounds like a lot, and it is still nothing on a 1 Gbps port.
What actually pushes the numbers up
Several things multiply the per-player figure, and they are the ones worth knowing about.
- Tick rate. The server sends a snapshot of the world several times per tick-worth of time. A 64-tick shooter sends roughly three times as much per player as a 20-tick sandbox for the same number of entities. See what is a server tick.
- Entity count. Every creature, dropped item, projectile, minecart and vehicle near a player has to be described to that player repeatedly. A giant mob farm or a base full of tamed animals raises bandwidth and CPU together. Reducing entity and mob lag helps both at once.
- View and simulation distance. A bigger radius means more chunks and more entities per player. This is the single biggest lever on Minecraft. See view and simulation distance.
- Exploration. Chunk data is the heavy part of Minecraft traffic. A player flying across unexplored terrain pulls megabytes; a player standing in a base pulls almost nothing. Pre-generating the world does not reduce the data sent but does take the generation work off the tick; see chunk loading and pre-generation.
- Resource packs and map downloads. A forced resource pack is downloaded by every joining player. A 200 MB pack and 100 joins a day is 20 GB. Host it off your server if you can; forcing a resource pack covers how. Counter-Strike 2 workshop maps and Rust custom maps work the same way: see Rust custom maps.
- Voice chat and live maps. Voice adds a steady stream per talking player in both directions (Simple Voice Chat setup), and a live map such as Dynmap serves tiles to web visitors from your server, which can easily exceed the game traffic if it gets popular (live maps with Dynmap).
Why upload matters more than download
A server is a broadcaster. It receives a small amount from each player (which keys they pressed, where they are aiming) and sends each player a description of everything happening around them. Send is many times larger than receive, and it scales with player count on both sides of the multiplication: more players means more recipients, and each recipient has more things to be told about.
This is the reason home hosting falls over long before a data centre does. A typical home connection has a fraction of its download speed available for upload, and a game server needs the upload. Sixteen players on a home Minecraft server can saturate a modest upstream, at which point everyone in the house notices. Hosting at home versus hosting with a provider goes through the full trade.
Why it is almost never the limit on a hosted server
The numbers above are the whole reason this guide exists. Data centre uplinks are measured in gigabits. A busy 60-player community server might peak at 20 or 30 Mbps. You are using a small fraction of one percent of the port.
In practice, the things that stop a hosted server scaling are, in order:
- Single-thread CPU, because the world's tick runs on one core. See why single-thread speed decides performance.
- RAM, when the world, the players and the mods do not fit. See how much RAM you need.
- Disk speed, during saves and chunk loads. See NVMe versus SSD versus HDD.
- Bandwidth, a distant fourth.
So when people say "our server cannot handle more players", it is almost never the network. How many players can my server handle breaks down what the real ceiling is.
Every GuavaHost server runs on a data centre connection with DDoS protection in front of it, and bandwidth is not something you size when you order: you choose RAM. If a flood of junk traffic ever does arrive, that is a different problem with a different answer, and how DDoS protection works covers it.
Measuring your own
You cannot read a bandwidth figure off the panel, because the Usage card tracks CPU, memory and storage. For a real number, check the game's own tooling (Minecraft server software with Spark can show network activity, and CS2 has net graph commands), and estimate your heaviest extras by hand: a forced resource pack times joins per day is usually the largest single item. If you are moving from a host that metered transfer, its last few invoices are your real baseline, and migrating from another host covers the move itself.
Common questions
Will more bandwidth reduce my server's lag?
Almost certainly not. Server lag means the world is overrunning its tick budget, which is a CPU problem, and player-side lag is usually ping, which is distance. Bandwidth only shows up as a problem when a connection is genuinely saturated, which is a home hosting symptom rather than a data centre one.
Does a bigger view distance really use that much more bandwidth?
Yes, because the area described to each player grows with the square of the radius. Going from 10 chunks to 16 roughly two and a half times the area, and both the chunk data and the entities inside it have to be sent. It also costs CPU, which is usually the reason to turn it down first.
Is a resource pack worth the transfer it costs?
Usually yes for the look of your server, but host the file somewhere else rather than serving it from your game server, and keep it as small as the art allows. Players download it once and cache it, so the cost is per new player rather than per join in most cases.
Do I get charged for bandwidth on GuavaHost?
Bandwidth is not something you pick when ordering a server, since the thing you size is RAM. If you have an unusual case, such as a very popular live map serving a lot of web traffic, mention it in Discord and the team can tell you whether it needs any thought.
Still stuck? Jump into our Discord at discord.gg/GuavaHost and the team will help you out.
- bandwidth
- networking
- upload
- players
- performance
Still need help? Our team is online 24/7 to answer questions.