Lab data

Modkeel lab data

The Modkeel lab runs mods in real Minecraft games, on servers and clients, and records what happened: which mods started, which ran together, which releases made for an older version did not start.

Lab data as of October 15, 2026.

By Minecraft version

By mod

15 mods have a page of their own: every mod the lab ran.

Ports

Community ports: mods built by Modkeel for Minecraft versions their authors have not released for yet, each tested in a real game.

How the frame rate benchmark works

How it was measured: a real Fabric client on a laptop with a GeForce GTX 1650 and a Core i7-10750H (an entry-level gaming laptop from 2020) joined the same world at 1280x720, render distance 24, VSync off. Each number is the median second of 20-second scenes over 2 to 27 runs. Runs without mods differed by up to 54%, so smaller changes are not counted. Frame rates depend on the computer: on a faster one the numbers here are usually higher. The percentage is the better guide, but it moves too: mods that cut rendering work (Sodium, More Culling) gain most when the graphics card is the limit, and mods that cut entities or game logic (Entity Culling, Lithium) when the processor is. Some effects belong to one GPU brand: Sodium's drop with mobs on 1.21.1 and 1.20.1 only happens on NVIDIA cards. Raw data: every run's frame rate per second, Minecraft and loader versions, GPU and driver, loaded mods and the SHA-1 of each jar, which Modrinth matches to the exact release.

How the server benchmark works

Each server optimizer runs alone on a fresh Fabric server with only Fabric API and Chunky, and the lab times the same scenes every run: how long the server takes to start, a tick with 400 mobs, a 64 by 64 redstone wire grid switched 100 times, and generating 960 by 960 blocks of new terrain, plus the memory in use after starting and after the new terrain. A mod counts as better or worse on a scene only when every one of its runs beat, or lost to, every run without it.

A plain server has none of the mods a big modpack loads, and no players: an optimizer that does nothing here can still help a busy server, and its cost here is what it adds to every server.

About this data

Loaded a world: a real Minecraft client started with the mod and joined a world. Server started: a real server finished starting with it. Benchmark: it ran through a timed lab scene. Did not start: the game stopped before it was playable.

In a grid of mods: World: loaded a world together on a client. Server: started a server together. Bench: ran together in a benchmark. Clash: did not run together. A dash: not tested together yet, which is not a conflict.

Each row is one run in the Modkeel lab, matched to the exact file that ran: other files may behave differently. A mod or pair missing here was not tested yet; that is not a conflict. Lab data is CC BY 4.0: attribute "Modkeel lab data (modkeel.com)". Ask the same from an AI assistant with Modkeel's MCP server, or let the Modkeel mod check your own game.