Calculator and measured guide, 2026
How much RAM does a Minecraft server need? We measured it.
Every Minecraft RAM calculator gives a different number per player, from 50 to 300 MB. So we ran a real server and measured.
We measured it: one player, three view distances
- View 6213 chunks352 MB
- View 10473 chunks424 MB
- View 16733 chunks576 MB
Empty server With one player
The short answer
For a small vanilla or Paper server, give Java 1 GB for up to 4 players, 2 GB up to 10 and 4 GB beyond. Add 1 to 1.5 GB for the system itself, so a 2 GB server suits a few friends and a 6 GB server handles 20.
View distance and modpacks change everything: in our test, one player at view distance 16 cost 224 MB more than at 6. Our free calculator adds it up, names your LiteKeep plan, and deploys it with DDoS filtering included.
Search “minecraft server RAM calculator” and you’ll get per-player figures anywhere from 50 MB to 300 MB. Most leave too little, or nothing, for the operating system, which is how servers end up crashing at peak time.
So we did the obvious thing. We ran a real Paper server, connected a test player, and measured the memory it actually used at three view distances.
The result surprised us. An empty server held about 220 MB of live data, yet the process showed over 2.5 GB in use. Java keeps whatever memory you hand it. That’s why “my server uses all its RAM” tells you almost nothing about what it needs.
The calculator below is built on those measurements. Set your players, view distance and modpack, and it gives you the Java memory, the start command and the plan to run it on.
In this guide you’ll learn
- A free calculator: Java memory, server memory, the start command and the plan
- What one player really costs, measured at view distance 6, 10 and 16
- Java memory vs server memory, and why they differ
- View distance: the setting that moves memory most
- Modpacks, plugins and Java versions
- How to measure your own server with spark
Chapter 1
The short answer
These are the Java memory figures for the game itself, at view distance 10 or less on a new world. Add 1 to 1.5 GB for the system and you have the server size:
| Players at once | Give Java | Whole server | LiteKeep plan |
|---|---|---|---|
| 1 to 4 | 1 GB | 2 GB | Lite 2 |
| 5 to 10 | 2 GB | 3 GB | Lite 6 |
| 11 to 20 | 4 GB | 5 GB | Lite 6 |
| Modpack | The pack’s own figure | Plus 1.5 GB | Use the calculator |
Chapter 2
Calculate your Minecraft server’s RAM
Move the sliders. The calculator starts from the player tiers, adds our measured cost of view distance and big player counts, takes your modpack’s own figure if it’s higher, adds room for the system, and names the Lite that fits.
Give Java
2 GB + 1 GB for the system = 3 GB server
java -Xms2G -Xmx2G -jar paper.jar --noguiPer-player and view-distance figures come from our own test server. Start here, then fine-tune with spark. On hourly billing a wrong first guess costs you an hour, not a month.
Try 8 players at view distance 16: the Java memory jumps from 2 GB to 3.5 GB. Then drop the view distance to 8 and watch it fall back.
Chapter 3
What one player really costs: our test
We ran Paper 26.1.2 with a fresh world each time and connected one test player, who stood at spawn for 100 seconds while chunks loaded. Then we forced a full garbage collection and read how much memory was really in use.
| View distance | Chunks sent | Empty server | With one player | One player cost |
|---|---|---|---|---|
| 6 | 213 | 219 MB | 352 MB | 133 MB |
| 10 (default) | 473 | 220 MB | 424 MB | 204 MB |
| 16 | 733 | 278 MB | 576 MB | 298 MB |
Two things stand out. First, view distance is expensive: going from 6 to 16 cost about 224 MB for one player, roughly 22 MB per step. Second, the chunk counts match the maths: at distance 10 the player received 473 chunks, close to the 441 in a square of (2×10+1)².
Chapter 4
Java memory vs server memory
A Minecraft server is a Java program. You tell Java how much memory it may use with -Xmx, and Java keeps what you give it.
In our test, with Java set to 4 GB, the empty server’s process showed 2.5 to 3.4 GB in use while its live data was about 220 MB. With one player, the process was over 3.9 GB. Tools like top show what Java reserved, not what the game needs.
That’s why the server needs more than Java’s share. Set Java’s memory about 1,000 to 1,500 MB below the server’s total, for example 6,500 MB on an 8 GB server, and keep -Xms and -Xmx the same.
# 8 GB server java -Xms6500M -Xmx6500M -jar paper.jar --nogui
- Don’t give Java all of the server’s RAM. Java needs memory outside its heap, and so does the system. On a 2 GB server, give Java 1 GB.
- Lag with free memory? Then the cause is usually processing time, not RAM. spark’s profiler shows you what, and a Lite with more vCPUs is the fix.
Chapter 5
View and simulation distance
Both settings live in server.properties. They range from 3 to 32, with 10 as the default.
- view-distance: how far players can see, in chunks. More chunks are loaded and sent to each player.
- simulation-distance: how far from a player mobs move and crops grow. It costs processing more than memory.
The area grows with the square of the distance: 441 chunks per player at 10, 1,089 at 16. That’s why our measured cost per player more than doubled from distance 6 to 16.
Chapter 6
Modpacks and plugins
Mods change memory more than anything else, and no general formula covers them. The modpack’s own recommendation is the best starting point.
For example, FTB StoneBlock 4 asks for at least 6 GB and recommends 8 GB. Treat that as the server’s Java memory and add the system headroom: about 9.5 GB in total, which is a Lite 16.
Some popular packs, like All the Mods 9 and 10, don’t give a figure. Start with 8 GB for Java on a Lite 16, then check with spark and adjust.
Plugins on Paper vary from nothing to a lot. Start at the player tier, then measure.
Chapter 7
Which Java version you need
Minecraft’s Java requirement has climbed with each big update. Get it wrong and the server won’t start.
| Minecraft version | Java |
|---|---|
| 1.17 | 16 |
| 1.18 to 1.20.4 | 17 |
| 1.20.5 to 1.21.x | 21 |
| 26.1 and newer | 25 |
Version 26.1 and its Paper builds need Java 25. Our test ran on Java 25.
Chapter 8
Measure your own server
Every estimate on this page is a starting point. spark, a free profiler plugin, shows your real numbers. Its documented commands include:
- /spark healthHow to check: Memory use, CPU and ticks per second at a glance.
- /spark gcmonitorHow to check: Watches garbage collections. Frequent long pauses often mean Java needs more memory.
- /spark heapsummaryHow to check: What is actually using the memory.
- /spark profilerHow to check: What is using the time, when the server lags with memory to spare.
Check memory at your busiest hour, not on a quiet evening.
Chapter 9
Hosting it on a VPS
A VPS keeps the world online when your PC is off, and doesn’t share your home connection. Three things matter.
- Location. Pick the country nearest the middle of your players. Our location pages show the minimum round trip to 26 cities.
- Size. Use the calculator. Hourly billing on our Lites means you can start small and move to a bigger Lite if spark says so.
- Protection. Game servers get attacked. Every Lite has DDoS filtering at the network edge, and the VPS security guide covers the server itself.
The default port is 25565. Open only that and SSH in your firewall.
Quick answers
Is 2 GB enough for a Minecraft server? For up to 4 players on vanilla or Paper at the default view distance, yes: 1 GB for Java, 1 GB for the system.
How much RAM for 20 players? About 4 GB for Java at view distance 10, so a 6 GB server. Raise view distance and it grows.
What about Bedrock? Bedrock’s dedicated server is a different program, and this calculator is for Java Edition. Bedrock runs on a Lite too: start small and check memory at your busiest hour.
Now it’s your move
Minecraft servers need less memory than most calculators suggest, until you raise the view distance or add a modpack. Then they need more than you’d guess.
Start with the calculator, deploy the Lite it names, and fine-tune with spark. With a Lite billed by the hour, a wrong first guess costs you an hour, not a month.