Vectors — Minecraft 1.21.1 Version Log

Loaders: Fabric and NeoForge. Install one loader build at a time. These are alpha builds; back up worlds before testing.

1.5.0-alpha.21 — October 17, 2026

• Added conservative opaque-terrain face filtering and retained terrain command reuse, with chunk mesh ownership checks that avoid capturing unrelated block meshes.
• Reused empty Fastutil list iterators and reduced temporary allocations in vanilla cubic sky/fog color sampling. Native sampler parity was checked across 520 cases, including fetch order.
• Passed local singleplayer resource reload, rendering, and cleanup checks on Fabric and NeoForge.
• Preliminary Fabric measurement, one completed run per scene: forest 3661 FPS, buildings 3121, glass/water 3730, chunk transitions 2888. Corresponding 1% lows: 511, 472, 499, and 284 FPS.
• A Sodium run was invalidated by a focus change, so this alpha has no valid new Sodium comparison. Treat these single-run figures as preliminary. The experimental terrain path requires -Dvectors.experimental1211Terrain=true.

1.5.0-alpha.19 — October 16, 2026

• Added conservative opaque-face filtering and retained terrain command reuse to the experimental terrain path.
• Fabric average FPS by scene (forest / buildings / glass-water / transitions): 3243 / 3331 / 3428 / 2856. Sodium in that campaign: 3098 / 2899 / 3100 / 2729.
• NeoForge average FPS by scene: 3359 / 3350 / 3466 / 2785. Sodium: 2894 / 3126 / 3497 / 2547.
• Two launches per configuration. Chunk-transition 1% lows were substantially below Sodium on both loaders; the averages do not mean Vector won every workload. See the linked full measurement reports in the project docs.

1.5.0-alpha.13 — October 16, 2026

• Added a 256-section terrain batch configuration to the experimental path.
• Fabric average FPS by scene (forest / buildings / glass-water / transitions): 3757 / 2803 / 4037 / 2624. Sodium in that campaign: 3218 / 3148 / 3568 / 2870.
• The 4037 FPS glass-water result is the earlier 4K-FPS run. Alpha.13 was slower than Sodium in buildings and transitions. Two candidate launches; results are scene-specific.
• NeoForge build is available, but this alpha's published comparison above is Fabric-only.

Notes

FPS measurements use separate local singleplayer processes with the same saved fixture and visible settings. Campaigns were run at different times; do not combine their numbers into one ranking. These are measured outcomes on one RTX 5070 / Core i5-14600KF system, not performance guarantees for other PCs.