Starlight is not a graphics-quality mod, so a single FPS number is a weak way to judge it. The project targets lighting work that can appear during chunk generation, block changes and server processing. Benchmark the workload the mod actually changes.

Upstream benchmark context: for the 1.20 generation, the project removed its old chunk-generation light-engine graph because the comparison was no longer useful, and noted that absolute client light-update times were unlikely to change FPS in most situations.

Measure lighting-heavy scenarios

Useful tests include repeated block-light changes, chunk generation, high-altitude block updates, busy server areas and world exploration. Keep the seed, render distance, simulation distance and mod list fixed.

Separate client rendering from server work

A GPU-bound client can show the same FPS even if server-side or chunk-side lighting work is reduced. Track frame-time spikes, tick stability, chunk completion time and responsiveness separately.

Use repeatable runs

Warm up the JVM, test the same route or scripted action several times, and compare medians rather than a single best result. Record the exact Minecraft, loader, Starlight and Java versions.

Change one variable at a time

Do not install Starlight, swap renderers, change Java flags and modify view distance in the same test. A controlled comparison is slower but produces a conclusion you can actually use.

Respect version-specific conclusions

Performance claims from 1.17 or 1.18 should not be copied directly into 1.20. Minecraft’s own engine changed over time, and the Starlight maintainer explicitly reconsidered the need for the mod on newer versions.

Know when not to use Starlight

If your profile is on a version the archived project never supported, or if another required mod conflicts with Starlight, the correct optimization decision may be to leave Starlight out rather than force it into the stack.

Related guides

Pair this with the common-errors guide and the Forge compatibility guide.