Sway: Obtain benchmarks

Created on 1 May 2018  路  3Comments  路  Source: swaywm/sway

  • CPU
  • RAM
  • Power

At the following loads:

  • Idle
  • Terminal running htop or something else that updates occasionally
  • Watching a video in mpv

We should compare 0.15 to 1.0, and 1.0 to KDE, Gnome

Most helpful comment

  • Latency (using https://github.com/mstoeckl/latencytool ), determined using a small window, and 187 fps camera. Results include display switching, camera, and USB latency, and as such are only suitable for relative comparisons. Latency values get worse with increased system load.

| Screen | GPU | Backend | Compositor | Application | Latency |
| ------------- | ------------- | ------------- | ------------- | ------------- |------------- |
| 60Hz IPS QHD | HD5500+i915 | X11/uxa | i3/none | latency_cv_qt | 25 ms |
| 60Hz IPS QHD | HD5500+i915 | X11/uxa | i3/none | latency_cv_xcb | 24 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | kwin | latency_cv_qt | 45 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | weston | latency_cv_qt | 41 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | weston | latency_cv_xcb | 41 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/X11 | sway | latency_cv_qt | 43 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_qt | 47 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_xcb | 44 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_wayland | 44 ms |

Note: latencytool was written yesterday; possibly more efficient frontends are in progress. Tests may be confounded by the terminal printing stdout. Weston lags when unrelated Xwayland applications are resized.

Second update: Implementation of xcb and Wayland frontends. Note that the Qt frontend uses QPainter on each frame, whereas the xcb/Wayland frontends do no rendering whatsoever. Replaced '50ms' measurement with wlroots/drm, sway, latency_cv_qt, with 47ms --it may be that DPI settings affect performance.

All 3 comments

  • Latency (using https://github.com/mstoeckl/latencytool ), determined using a small window, and 187 fps camera. Results include display switching, camera, and USB latency, and as such are only suitable for relative comparisons. Latency values get worse with increased system load.

| Screen | GPU | Backend | Compositor | Application | Latency |
| ------------- | ------------- | ------------- | ------------- | ------------- |------------- |
| 60Hz IPS QHD | HD5500+i915 | X11/uxa | i3/none | latency_cv_qt | 25 ms |
| 60Hz IPS QHD | HD5500+i915 | X11/uxa | i3/none | latency_cv_xcb | 24 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | kwin | latency_cv_qt | 45 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | weston | latency_cv_qt | 41 ms |
| 60Hz IPS QHD | HD5500+i915 | drm | weston | latency_cv_xcb | 41 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/X11 | sway | latency_cv_qt | 43 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_qt | 47 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_xcb | 44 ms |
| 60Hz IPS QHD | HD5500+i915 | wlroots/drm | sway | latency_cv_wayland | 44 ms |

Note: latencytool was written yesterday; possibly more efficient frontends are in progress. Tests may be confounded by the terminal printing stdout. Weston lags when unrelated Xwayland applications are resized.

Second update: Implementation of xcb and Wayland frontends. Note that the Qt frontend uses QPainter on each frame, whereas the xcb/Wayland frontends do no rendering whatsoever. Replaced '50ms' measurement with wlroots/drm, sway, latency_cv_qt, with 47ms --it may be that DPI settings affect performance.

What does "wlroots/EGL" mean? Is this DRM?

Yes, updated.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

WhyNotHugo picture WhyNotHugo  路  3Comments

mcmfb picture mcmfb  路  3Comments

StephenBrown2 picture StephenBrown2  路  4Comments

soymjolk picture soymjolk  路  3Comments

Olie440 picture Olie440  路  4Comments