BETAMCC
Performance

The fastest
of them all. Honestly.

"Honestly" means: only numbers we measured ourselves, with hardware, scene and settings next to them. We don't publish numbers for other clients we haven't measured.

Test setup

Client MCC BETA · native 26.3

API Vulkan (NVRHI)

Resolution 1920×1080

Render distance 12 chunks

Scene mob zoo: 60+ mobs, block entities, terrain

Mode --bench 5, uncapped, release build

GPU: NVIDIA GeForce RTX 5080 Laptop

CPU: AMD Ryzen 9 9955HX3D

OS: Linux (headless gamescope)

Server: our local test server

Frame rate per preset

From vanilla look to everything on.

Off
4,755 fps
Low
1,166 fps
Medium
479 fps
High
370 fps
Ultra
90 fps
PresetfpsGPU ms / frameAdds
Off4,7550.14vanilla look
Low1,1660.78shadows, bloom, waving plants
Medium4792.06SSAO, screen-space GI, god rays, reflections
High3702.68coloured shadows, voxel GI, smooth water
Ultra9011.2192-block shadows, full-res GI, 2 rays, 3 bounces

Re-measured on 2026-10-23 with a debug-info build, same scene: High 339 fps / 2.91 ms GPU, Ultra 83 fps. In game at 1080p High with the perf overlay on: 355 fps.

Vanilla look

fps on the terrain view at render distance 8: 0.058 ms of GPU time per frame.

  • 10–13kfps at 1080p, RD 8 and 12, motion blur off
  • ~120 msto bake all 35,723 block states
  • ~75 µsto mesh a surface chunk section

Same laptop, renderer test world and spawn view; the 12.8k figure in a 1278×1576 window, immediate present.

Pacing

Smooth, not just fast.

High fps means little if frames arrive unevenly. Present intervals we measured with frame generation (FG) and LateWarp at 240 Hz:

CaseRendered fpsShown / sInterval mean / stddev
FG off, cap 60606016.67 / 0.26 ms
FG 2×, cap 60, FIFO601208.33 / 0.27 ms
FG 2×, cap 1201202404.17 / 0.33 ms
LateWarp, cap 60602364.24 / 0.35 ms
FG off, uncapped8,7702404.17 / 0.15 ms

RTX 5080 Laptop GPU, demo scene, 1278×1576 window. A generated frame costs 0.5–0.9 ms of GPU time.

Why

Where the speed comes from.

  1. Zero-hitch rules

    No heap allocation, blocking I/O or pipeline creation on the render thread in steady state.

  2. Decoupled present

    A present thread paces frames and draws straight to the back buffer.

  3. Threaded meshing

    Chunk sections mesh on a job system across all cores, never on the main thread.

  4. Cached UI layer

    The HUD is rebuilt at a capped rate and composited instead of redrawn every frame.

  5. Measured every change

    GPU timers per pass and a built-in bench mode.

Goal for 1.0: at least 10,000 fps at render distance 8, low settings, 1080p, on high-end hardware. A goal, not a result yet.