Skip to main content

First-party measurements · controlled baseline

Leaf vs Paper vs Leaves in a direct benchmark

In this synthetic test series, Leaf reached 20 TPS at 1.9 ms median tick time, Paper reached 20 TPS at 1.0 ms, and Leaves reached 20 TPS at 2.2 ms. This small controlled sample documents the workload and does not establish an overall winner.

Dataset published on 2026-10-12 and last modified on 2026-10-12.

Concise finding, methodology and limits

This measurement compares Leaf, Paper and Leaves under identical conditions on the same hosting node. Every server received 4 GB RAM and 100% CPU. For each fork, 3 complete runs used the same seed, 4225 pregenerated chunks, 200 cows and 5 plugins. Leaf reached 20 TPS at 1.9 ms median tick time, Paper reached 20 TPS at 1.0 ms, and Leaves reached 20 TPS at 2.2 ms. The TPS values show whether the normal Minecraft ceiling of 20 TPS was maintained under this workload. The measured MSPT values describe only this synthetic workload; small differences in this sample do not establish an overall performance ranking. Process CPU is a measured game-process value, not the purchased CPU limit. The test did not cover live players, network latency, long-term stability, different worlds or other plugin configurations. Updated on 2026-10-12, the public dataset contains aggregate values only and links the reproducible methodology and self-hosted result files. Open the dataset and methodology.

Controlled test results

Median milliseconds per tick
Comparison of median tick time for Leaf, Paper and Leaves. Lower values only mean less measured tick processing time within this test series.

The shared scale starts at zero. The chart may be shared; use the dataset and limitations with it for context.

Open benchmark chart as SVG
Server software Usable runs Median TPS Median MSPT Median process CPU Measured
Leaf 3/3 20 1.9 ms 5.825%
Paper 3/3 20 1.0 ms 3.340%
Leaves 3/3 20 2.2 ms 6.184%

What the result means

  • Median TPS was 20 for Leaf, 20 for Paper and 20 for Leaves. Minecraft’s normal ceiling is 20 TPS.
  • Median tick time was 1.9 ms for Leaf, 1.0 ms for Paper and 2.2 ms for Leaves. This small sample does not support a general performance ranking.
  • Median process CPU was 5.825% for Leaf, 3.340% for Paper and 6.184% for Leaves. Different worlds, plugins, player counts and settings can change the relationship.

Test setup

Every project used the same seed, 4225 pregenerated chunks, 200 cows and 5 plugins. Each server had 4 GB RAM and 100% CPU. A 120-second warm-up preceded 300 seconds of measurement sampled every 15 seconds. Usable runs: Leaf: 3; Paper: 3; Leaves: 3.

Measurement limits

  • The workload is synthetic and does not represent live players or one specific production setup.
  • The test does not measure player latency, network quality, plugin compatibility, support quality or long-duration stability.
  • Choose a fork for your own world, required plugins and desired vanilla behavior, then test it with the workload you expect.

Earlier Leaves aborts were traced to a race condition in the benchmark tool, not a Leaves defect. After the readiness check was fixed, all five runs of the verification campaign completed in full; the table counts the three runs of the 20 Sept 2026 measurement.

Project sources

Use the respective project pages for downloads, version notes and configuration details:

Download aggregate benchmark data as JSON

The public file contains aggregate values only. Server and panel identifiers, addresses, raw console output and individual samples are excluded.

More game-serverhosting product facts Configure a Minecraft server