Homemade Latency Rig Debunks Linux Gaming Lore: Wayland Isn't the Problem

Tired of unverifiable folk wisdom about optimizing Linux for competitive gaming, the author built a hardware latency tester to settle the debate empirically. The rig uses an Adafruit QT Py RP2040 that acts as a 1000 Hz USB HID mouse: the instant it fires a click, it begins sampling a photodiode taped to the screen every ~24 µs, streaming 12,000 samples per click to a host tool that pinpoints the first brightness change and computes true end-to-end system latency. Testing ran on a static, CPU-bound scene in Diabotical (DirectX 11 via Proton), with 100 clicks per configuration.

The results deflate several popular claims. X11 does beat Wayland, but only by 0.14–0.22 ms per scenario — nowhere near enough to justify Wayland’s reputation for feeling ‘off.’ The real penalty comes from XWayland, which adds a hefty 3.13 ms over native Wayland. Variable Refresh Rate turned out to be the single most impactful setting, cutting 0.26–0.45 ms and noticeably tightening the latency distribution because frames scan out as soon as they’re ready. The dxvk-low-latency frame pacer shaved about 0.20 ms in capped tests and up to 0.84 ms in uncapped scenarios where its ability to absorb frame-time spikes and prevent render-queue buildup actually matters.

The broader takeaway is that most of the tweaks gamers obsess over produce differences smaller than 1 ms — well within the range of placebo — while the choices that genuinely move the needle are enabling VRR and, above all, running native Wayland instead of XWayland. The project also doubles as an accessible open-hardware effort, drawing on OSLTT, m2p-latency, and Open-Source-LDAT designs for anyone wanting to measure rather than guess.