Complementary Shaders Performance Optimization Guide

Most shader performance advice is a pile of tips. This guide is a method instead: a repeatable way to find out what Complementary Shaders costs on your machine, in your world, and to buy back the frames you are owed without giving up the look you installed the pack for. Follow it once, properly, and you will not need to re-tune blindly ever again.

The principle: measure, change one thing, measure again

Shader cost is scene-dependent. A setting that is free in a plains biome is expensive in a jungle; a change that wins frames at noon may do nothing at sunset. That is why copied settings disappoint: they encode someone else's scenes. The method below replaces copying with testing, it takes about twenty minutes and produces a configuration that is genuinely yours.

Step 1: Choose your test scenes

Pick three places in your actual world and note how to return to each:

• Your busiest scene: main base, storage, farms, entities, the worst case you actually play in.
• An open vista: water, distance, sky, where render distance, reflections and clouds are taxed.
• An underground or night scene: where lighting and shadows do their work.

For each test, stand in the same spot, look the same direction, and set the same time of day. Consistency is what makes your numbers comparable.

Step 2: Record a baseline

With your current settings, visit each scene and note the frame rate after it settles (give it twenty to thirty seconds, the first moments after loading are unrepresentative). Note the average and, just as importantly, whether it feels smooth: a steady 45 often plays better than a jittery 60. If your problem is jitter rather than low averages, your diagnosis is different, see lag and stuttering.

Step 3: Work the levers in order of leverage

1. Quality profile, one step at a time (presets explained).
2. Render distance, the multiplier on everything (dedicated guide).
3. Shadow distance, then resolution (shadow settings).
4. Cloud quality, the constant sky tax.
5. Reflection quality, scene-dependent; huge near water, free in a desert.
6. Resolution, the global dial, used deliberately.

After each single change, revisit all three scenes and record. Stop changing a lever when its cost in looks exceeds its gain in frames for you. The quick-fix summary of this same order lives in how to increase FPS.

Step 4: Fix the environment, not just the pack

• Loader stack: Iris + Sodium returns real budget (compatibility guide).
• GPU selection and drivers: the game must run on your dedicated GPU with current drivers (GPU guide).
• Memory allocation: modded setups need enough RAM allocated; starvation stutters independently of graphics.
• Background competition: recording, streaming and browsers share your GPU. Test with your real background load, then decide what you are willing to close.

Step 5: Write it down

Record your final configuration, profile, render distance, and every override, somewhere outside the game. Settings files are per version filename and can reset on updates; your note is the insurance. The update routine is in updating without losing settings. Re-run this method only when something material changes: new GPU, new Minecraft version, a major pack update, or a move to a much heavier world.

Reading frame-time behaviour without special tools

The baseline step asks you to note whether play feels smooth, and that judgement can be made more precise without any software. Watch how the frame rate behaves rather than only what it averages. A counter that sits in a narrow band while you turn slowly is a healthy sign, even if the number itself is modest. A counter that swings widely in the same view, dipping when you face one direction, recovering when you face another, tells you a particular system is taxing that view: water in frame points to reflections, dense canopy to shadows, open sky to clouds and atmosphere. A rate that is steady in place but lurches while travelling points away from the pack's effects and towards chunk loading. A rate that starts well and erodes across a session points towards heat or memory pressure. Recording which pattern you saw beside each test scene turns vague impressions into a diagnosis, and it is the difference between tuning by evidence and tuning by hope.

Adjusting the method for multiplayer

The three-scene method assumes you control the world, and servers remove some of that control. You cannot set the time of day, pause entity activity, or guarantee the same neighbours are online, and your effective view is capped by the server's own view distance regardless of your client setting. Adapt rather than abandon the method: choose a quiet moment on the server, return to the same landmark and view direction for each test, and repeat a measurement if a crowd passes through mid-test. Treat results as comparisons between your own settings on that server, not as absolutes, a hub scene with dozens of players is a stress test, not a baseline. If a configuration works in your single-player test scenes but struggles only in crowded server areas, the cause is entity concentration in view, and profile trims aimed at sky or water will do little; shorter render distance and acceptance that hubs are exceptional scenes is the more honest response.

Keeping a tuning log you will actually use

Step 5 asks for a written record, and the format matters less than a few habits that keep it useful. Date each entry, because a configuration tuned for one pack version or game version is a historical document once those change. Change the log only when you change settings in the game, a log that drifts from reality is worse than none. Note the scene that forced each decision ("dropped shadow distance because the jungle base dipped") rather than only the value; months later, the reason tells you whether the compromise is still necessary after a world move or hardware change. Keep the log where you keep your other game notes, and revisit it only on the triggers already listed, new hardware, a major pack or game update, or a substantially heavier world. Players who log this way typically tune once, play for months, and re-tune in minutes; players who tune from memory repeat the same experiments every update.

Frequently asked questions