Minecraft Server Performance Benchmarks 2026: Vanilla vs Paper vs Folia

Choosing the right server software for Minecraft directly impacts performance, stability, and player experience. In 2026, we have new server core versions, Java updates, and JVM optimization techniques. We ran comprehensive benchmarks and are ready to share the results.

Benchmark Methodology

We standardized all conditions for objectivity:

• Profiler: Spark Profiler v1.10 for monitoring TPS, MSPT, and memory usage
• Player simulation: 25, 50, 100, and 200 bots via Headless MC
• World: pre-generated 5000x5000 block world with villages and farms
• Plugins: standard set (EssentialsX, LuckPerms, WorldGuard) - 12 total
• Hardware: AMD Ryzen 9 7950X, 64GB DDR5-5600, NVMe Gen4
• Network: 1 Gbps dedicated, filtered through MineGuard

Each test was run three times for 30 minutes. We recorded average values and peak loads.

Server Software Comparison: TPS Under Load

TPS (ticks per second) is the primary performance metric. The ideal value is 20.0 TPS.

Software	25 Players	50 Players	100 Players	200 Players
Vanilla 1.21.4	19.8	16.2	7.0	2.1
Spigot 1.21.4	20.0	19.1	14.0	8.3
Paper 1.21.4	20.0	19.9	18.0	14.7
Purpur 1.21.4	20.0	20.0	19.0	15.9
Folia 1.21.4	20.0	20.0	20.0	18.4

Key finding: Vanilla drops below 20 TPS at just 50 players. Paper and Purpur hold stable 18-19 TPS at 100 players. Folia, with its multi-threaded region processing, shows the best results under heavy load but requires compatible plugins.

RAM Usage Comparison

Memory consumption with standard 8 GB allocation (-Xmx8G):

Software	25 Players	50 Players	100 Players	200 Players
Vanilla	3.2 GB	4.8 GB	7.1 GB	OOM
Spigot	2.8 GB	4.1 GB	5.9 GB	7.6 GB
Paper	2.4 GB	3.5 GB	5.1 GB	6.8 GB
Purpur	2.5 GB	3.6 GB	5.2 GB	6.9 GB
Folia	2.9 GB	4.0 GB	5.8 GB	7.4 GB

Paper manages memory most efficiently thanks to chunk and entity ticking optimizations. Folia uses slightly more due to multi-threading overhead, but this is justified by its stable TPS.

Chunk Loading Speed

Test: teleport 10,000 blocks and measure full chunk loading time at view-distance 10:

Software	Load Time (sec)	Chunks/sec
Vanilla	8.4	74
Spigot	5.1	122
Paper	2.3	271
Purpur	2.1	297
Folia	1.4	446

Async chunk loading in Paper/Purpur provides a 3-4x advantage over Vanilla. Folia loads chunks in parallel across regions, nearly doubling Paper's speed.

Java Version Impact

Testing Paper 1.21.4 with 100 players on different JDK versions:

JDK	TPS	MSPT (avg)	RAM	Startup (sec)
OpenJDK 17	17.2	42ms	5.4 GB	12.1
OpenJDK 21	18.0	38ms	5.1 GB	10.8
GraalVM CE 21	18.6	35ms	4.8 GB	9.4
GraalVM EE 21	19.1	32ms	4.6 GB	8.7

Upgrading from Java 17 to Java 21 yields a 5-8% TPS improvement. GraalVM Enterprise delivers the best results thanks to its advanced JIT compiler, though the paid license is not for everyone.

JVM Flags Optimization

Comparing popular launch configurations (Paper 1.21.4, Java 21, 100 players):

Configuration	TPS	GC Pause (avg)	GC Pause (max)
Default flags	16.8	45ms	210ms
Aikar's Flags	18.0	18ms	85ms
Aikar's + ZGC	18.3	8ms	22ms
GraalVM + Aikar's	18.9	12ms	45ms

Recommended Aikar's Flags for Java 21:

java -Xms8G -Xmx8G -XX:+UseG1GC -XX:+ParallelRefProcEnabled \
-XX:MaxGCPauseMillis=200 -XX:+UnlockExperimentalVMOptions \
-XX:+DisableExplicitGC -XX:+AlwaysPreTouch \
-XX:G1NewSizePercent=30 -XX:G1MaxNewSizePercent=40 \
-XX:G1HeapRegionSize=8M -XX:G1ReservePercent=20 \
-XX:G1MixedGCCountTarget=4 -XX:InitiatingHeapOccupancyPercent=15 \
-XX:G1MixedGCLiveThresholdPercent=90 \
-XX:G1RSetUpdatingPauseTimePercent=5 \
-XX:SurvivorRatio=32 -XX:+PerfDisableSharedMem \
-XX:MaxTenuringThreshold=1 -jar server.jar --nogui

Hosting Comparison: Dedicated vs VPS vs Shared

Parameter	Dedicated	VPS (KVM)	Shared
TPS @ 100 players	18.0	15.2	11.4
MSPT stability	High	Medium	Low
Latency spikes	Rare	Frequent	Constant
Price/month	$50-150	$15-40	$5-15
Customization	Full	High	Limited
DDoS protection	Varies	Basic	Shared

Shared hosting suits servers with up to 20 players. VPS handles up to 50. For 100+ players, we recommend a dedicated server with single-core frequency of 4.5 GHz or higher.

Single-Thread vs Multi-Thread Performance

Minecraft on traditional server cores (Vanilla, Spigot, Paper, Purpur) runs primarily on a single thread. Therefore:

• Most important: single-core clock speed (4.5+ GHz recommended)
• Less important: core count (4-6 cores sufficient for Paper)
• Exception: Folia utilizes multiple cores, and 8+ cores provide noticeable gains

CPU ranking by single-thread performance for Minecraft:

CPU	Single-thread Score	TPS @ 100 (Paper)
AMD Ryzen 9 7950X	2050	18.0
Intel i9-14900K	2100	18.3
AMD Ryzen 7 5800X	1600	16.5
Intel i7-12700K	1750	17.1
AMD Ryzen 5 3600	1300	14.2

Network Bandwidth Requirements

Real traffic consumption per player depends on activity:

Players	Outbound Traffic	Peak Traffic	Recommended Link
25	15 Mbps	25 Mbps	100 Mbps
50	35 Mbps	55 Mbps	200 Mbps
100	80 Mbps	120 Mbps	500 Mbps
200	170 Mbps	260 Mbps	1 Gbps

Beyond bandwidth, network stability and DDoS protection are critical. During an attack, even a perfectly optimized server becomes unreachable.

Network Performance and Protection

Server optimization is only half the equation. The network layer matters just as much. At MineGuard, we built a proxy filter that adds less than 1 ms of latency while filtering attacks. Server-side optimization combined with network protection delivers the best experience for your players.

Our tests showed that during a 50 Gbps DDoS attack, an unprotected server loses 100% of its players. With MineGuard protection, players do not even notice the attack, and TPS stays at 20.0.

Final Recommendations

• Up to 30 players: Paper + Java 21 + Aikar's Flags + VPS with good CPU
• 30-100 players: Paper/Purpur + Java 21 + Aikar's Flags + dedicated server
• 100+ players: Folia (if plugins are compatible) or Purpur + GraalVM + powerful dedicated
• Always: DDoS protection for stable operation even under attack

Choosing the right server core, Java version, and JVM flags can improve performance by 50-100% without any extra hardware spending.