Download File

# Experimental C2ME edition for Minecraft 26.3 **EXPERIMENTAL ALPHA: a one-off C2ME variant of Smooth Server.** C2ME is a development build; this is an exception to the regular pack's stable-mod selection. Test on a copy of your world and keep backups before using it on a live server. For crashes, compatibility problems or other issues, email **[myuui@myuui.com](mailto:myuui@myuui.com)**. Include Minecraft 26.3, the experimental pack version, steps to reproduce, and relevant server logs or crash reports. ## Changes - Added [C2ME 0.4.2-alpha.0.90+26.3](https://modrinth.com/mod/c2me-fabric/version/ODMLK8M9) (alpha). - Based on the published **26.3 Fabric-SmoothS v1** pack; the other 15 mod files and supplied runtime configuration remain unchanged. - C2ME automatically disables four conflicting ModernFix options: `release_protochunks`, `paper_chunk_patches`, `cache_strongholds`, and `chunk_deadlock`. No whole mods were removed. - Requires **Fabric Loader 0.19.5** and **Java 25 or newer**. Install on the dedicated server; clients do not need this pack. - Still missing **Async, Immersive Optimization and ScalableLux**. ## Five paired terrain tests **Five fresh runs of default Smooth Server versus five fresh runs of Smooth Server + C2ME: 10 measured server runs.** Order alternated within each pair, with one server running at a time. Each trial cloned the same validated world template and matched its initial hash. The earlier standalone run and historical default-pack runs were excluded. | Median measurement | Default Smooth Server | Smooth Server + C2ME | Reduction | Paired outcome | |---|---:|---:|---:|---| | Terrain loading/preparation (s) | 13.389 | 9.233 | 31.0% | All 5 pairs improved | | Terrain CPU time (CPU s) | 51.609 | 44.484 | 13.8% | All 5 pairs improved | | Startup (s) | 9.203 | 8.891 | 3.4% | Mixed / inconclusive | | Maximum sampled batch-end resident memory (MiB) | 1121.984 | 1086.355 | 3.2% | Mixed / inconclusive | **Lower is better.** Percentages above compare medians. C2ME reduced terrain loading/preparation time and terrain CPU time in every pair. Startup and sampled memory improved in three pairs and regressed in two, so those results are mixed. | Sample pair | Default terrain time (s) | C2ME terrain time (s) | Paired time reduction | |---|---:|---:|---:| | 1 | 11.874 | 8.674 | 26.9% | | 2 | 13.580 | 9.437 | 30.5% | | 3 | 13.389 | 9.079 | 32.2% | | 4 | 13.563 | 9.233 | 31.9% | | 5 | 13.295 | 9.283 | 30.2% | Default terrain times ranged from **11.874 to 13.580 s**; C2ME ranged from **8.674 to 9.437 s**. The median paired time reduction was **30.5%**. ### Method and limits Each run loaded and prepared **512 fresh target chunks in eight sequential batches of 64**, after generation warmup in a separate unmeasured region. Simulation was frozen. Every target chunk was checked for loading completion; neighboring chunks may also be generated. Forced chunks were released between batches, with saves outside the timed measurement. Local shared Windows 10 Pro host; Temurin Java 25.0.4; 2 GiB maximum heap; four processors visible to the JVM; 4 GiB process memory limit; 50% host CPU cap; seed 623498172; no players; warm disk caches. C2ME used its default three-worker executor. Mod hashes, initial world hashes, supplied configuration, status responses and clean shutdowns were validated. All five pairs passed on their first attempts. Memory is the largest of eight batch-end resident-process samples, not a continuously measured peak. The results apply to this terrain workload and the complete C2ME variant, including its automatic ModernFix compatibility adjustments. Five improving pairs describe consistency in these trials, not statistical confidence or a guarantee for other hardware, worlds or player loads. Tick performance, FPS, network throughput and player capacity were not measured here. Benchmark reference: `c2me-paired-20261011T221322Z-11984`.