Measuring Input Latency on Linux: X11 vs. Wayland, VRR, and DXVK
Points and comments are a snapshot, not live.
Hardware latency measurements show Wayland is only 0.2ms slower than X11, but XWayland adds 3.1ms.
The author built a custom photodiode-based latency measurement device to compare end-to-end system latency on Linux. Testing a competitive FPS game (Diabotical) at 500Hz on an RTX 4070 SUPER, they found that native Wayland adds only 0.14-0.22ms over X11. XWayland, however, adds 3.13ms. VRR provides the largest single improvement (0.26-0.45ms), and the dxvk-low-latency framepacer yields a consistent 0.10-0.29ms gain in capped scenarios, with a 0.84ms gain when uncapped. The fastest configuration (X11, VRR, low-latency) achieved 4.21ms median latency.
What commenters are saying
Commenters praised the article, with a top comment noting that the XWayland result likely explains why people perceive Wayland as laggy: they were running X11 games through XWayland. A counterpoint argued that testing only KDE Plasma's compositor cannot generalize to all Wayland compositors, as they differ greatly in implementation. Several commenters expressed frustration with the Wayland transition, citing broken features and wasted volunteer effort. Another noted that at 500Hz the differences are small, but at 60Hz the impact of VRR and consistent pacing becomes much more significant.