Minecraft FPS Lab

Each experiment uses three ABBA blocks: twelve fresh client launches and six paired comparisons. Each launch excludes its first complete conditioning capture. The remaining capture includes three measured scene passes, with a 30-second warmup before each. Passes are aggregated per launch.

Lower P99 and P99.9 frametimes mean smoother delivery; higher FPS means greater throughput. Changes compare B with A. A 95% paired bootstrap interval and 3% practical threshold determine the provisional outcome. Unstable controls and mixed evidence remain inconclusive. These exploratory results have no correction for testing many claims. A promising result needs an independent session and another scene before joining the recommended stack.

The mod's elapsed-time 1% and 0.1% lows differ from frame-count percentiles. Exact settings, mod hashes, Java builds, collector names, raw frames, world hashes and rejected attempts stay in the private SSH archive. This public snapshot refreshes every 30 seconds.

Visual and view-distance changes are quality/performance tradeoffs. Findings apply to this hardware, scene and baseline. The campaign excludes Windows, driver, BIOS, power-plan and network tuning.