Bons and Furious 1.0.26

More performance on top of the performance mods you already run. 127 controls for Minecraft 1.20.1, Forge and 44 optional mod targets, one JAR for client and server.

Latest whole-modpack benchmark

Both arms keep ModernFix, FerriteCore, Radium and C2ME. The client also keeps Embeddium, ImmediatelyFast and EntityCulling, plus Oculus, Distant Horizons, Complementary shaders and Fresh Animations. Four matched runs per arm, same PC and seed, 2 October 2026; C2ME reduceLockRadius=false throughout.

• 5.6× world-generation throughput: 512 new chunks in 102 vs 567 seconds.
• 4.8× faster spawn preparation: 58 vs 277 seconds in a new singleplayer world.
• 50% less time from game launch to standing in a new world: 242 vs 482 seconds.
• 81% less CPU and 87% less allocation per generated chunk.
• 4.2× as many Distant Horizons LOD records built in the same four-minute session.
• Client start-up 128 vs 136 seconds; server start-up unchanged. Server live heap after start-up +70 MB (2.30 vs 2.23 GB).

Measured FPS gains from the earlier 1.0.21 benchmark (30 September): +10% average FPS and +16% 1% lows with 64 animated mobs; +29% 1% lows in the GPU-bound overlook, whose average FPS was unchanged. The existing client performance stack was present in both arms. These are historical measurements, not extrapolated 1.0.26 results. The latest saved-world absent arm froze the integrated server; the wiki reports that stability result separately and excludes its FPS claims.

Twenty controls since the public 1.0.23 release

• Embeddium + Fusion: hidden connected-texture faces skipped before quad construction, section-cache expiry from the oldest end, weighted-model picks from a table. Full section rebuild 5.2 → 3.9 seconds, 41% less allocation for the hidden-face control, byte-identical mesh output in the tested scenes.
• Farmer's Delight: one tool-action registry scan per action and recipe load. Recipe packet decoding 8.8 → 0.4 seconds; client recipe rebuild 5.5–6.4 → 1.4 seconds. Every ingredient retains its own stacks and the same item order.
• Entity Model Features / Fresh Animations: variable indexes and model-part lookup memoization; entity-renderer rebuild 15.1–16.6 → 8.8 seconds.
• JEI: ingredient stack cache and item-registry wrapper reuse; the same expiry semantics, categories, recipe content and stack content in the checks.
• Distant Horizons: wait instead of spinning while queued updates debounce; direct saved-chunk reads restored when the conflicting C2ME IO modules are off.
• Relics: foreign-capability queries return the same empty result without constructing a discarded LazyOptional (128 → 0 bytes per query).
• Fusion: cheaper connection-key hashing, small block-set checks and chunk layer lookup. These isolated method savings were below in-game timing noise.
• Minecraft and Radium: density-wrapper table sizing, chunk-status names, per-pass camera-fluid memoization, Forge-safe block-entity state and lazy POI section lookup, selected Radium ticket/spawning options with explicit settings respected.
• Hostile Villages: queued village mobs wait for their chunks to finish, avoiding a reproduced 8.4–17-second server tick; the patched reproduction stayed below 47 ms. Queue order and mob setup are retained, with documented timing differences.

The release includes 1.0.24 and 1.0.25, which were not published separately. The separate BonsFusion alpha.2 changes are not included in these release claims. Per-patch figures describe the named workload and do not add together.

Install and verification

Replace the older Bons and Furious JAR with bons_and_furious-1.0.26.jar and restart. Minecraft 1.20.1, Forge 47.3.22 or newer within 47.x; tested with 47.4.16. No required mod dependencies. Existing config choices are retained; missing keys default to true. Every control has its own switch in config/bons_and_furious.properties. Unsupported target bytecode is skipped with a log line.

The exact published JAR passed the final client/server qualification and the latest benchmark. Its SHA-256 is 197c939fe6d9d1fa67c05980db327d21eafd371735236ac2b060b57728f78efa.

All 127 controls · Benchmark and method · Source