Performance hit every X seconds when playing games on high resolution/quality

[greenfoo]

Hyprland version

git master from August 8th 2025 (b619f39)
Aquamarine 0.9.4

Regression

Not sure. I have tried with older versions and the issue is still there... I will keep trying with even older versions

Describe the bug

While playing games, if a second monitor is connected, FPS drop to 0-5 every X seconds for a short period of time.

Reproduction steps

I'm still trying with different configurations to make it more general but, for reference, this is how I originally reproduced the issue:

1. On an Intel Core Ultra 165U laptop, start the flatpak version of Steam.
2. Run "Magic: The Gathering Arena". Keep it open on the lobby.
3. Monitor FPS and make sure they are stable.
4. Connect an external monitor over USB-c
5. Now, every X seconds FPS drop to 0-5 for 10~20 seconds.

Some notes:

• "X" changes on every test run I make, but for one particular test run, "X" is extremely stable. Example:
  • Test run number 1: X is exactly 110 seconds. Every 110 seconds, and for 20 seconds, the FPS drop
  • Test run number 2: X is exactly 51 seconds. Every 51 seconds, and for 16 seconds, the FPS drop
  • etc...
• The game does not need to be "Magic: The Gathering" (but I included it because it is free). It also happens with others, such as "DayZ". I thought it might have been related to Wine, but I found a game ("Axiom Verge") which does not suffer from this even when running under Wine/Proton. I suspect that it has to be a game which heavily uses the GPU, but I'll keep investigating.
• I tried with both the "i915" and "xe" drivers and the issue does not go away.
• In the 10~20 seconds when the FPS problem is happening, the CPU usage of the Hyprland process spikes from ~10% to ~20%. Once the FPS are back to normal values the Hyprland process goes back to ~10%.
• I set debug:disable_logs = false but nothing happens when the FPS go down. Is there any other option to increase the logging level?
• Is there some timer of timer / garbage collector in Hyprland that is only triggered when a second monitor is plugged in?
• What else can I do to obtain more data?

System info and config

systeminfo.txt

[greenfoo]

More data:
I ran strace on the Hyprland process when it is working fine (strace_ok.txt) and when the FPS drop to almost zero (strace_ko.txt) and in the second case I observe many more calls to epoll_wait():

strace_ok.txt
strace_ko.txt

[vaxerski]

explicit sync fail? Dk, can you share a hyprland log from when this happens?

[greenfoo]

I left a terminal open on one side of the screen and the game running on the other side, and there was no output being generated by hyprland when the issue took place:

[LOG] cursorImage request: shape 1 -> default
[LOG] CWLSurface 55883f70f6e0 called destroy()
[LOG] [AQ] drm: Cursor buffer imported into KMS with id 302
[LOG] Hyprctl: new connection from pid 73963
[LOG] [CXDGSurfaceResource] xdg_surface 55883fee4440 requests geometry 0x0 931x1116
[LOG] [CXDGSurfaceResource] xdg_surface 55883fee4440 requests geometry 0x0 931x1116
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] [AQ] [libinput] event5  - palm: keyboard timeout

[LOG] Keybind triggered, calling dispatcher (8, , 65289, workspace)
[LOG] Searching for matching rules for foot (title: foot)
[LOG] [CWLDataDeviceProtocol] New X11 offer 5588403e95c0 for data source 558840420bc0
[LOG] Set keyboard focus to surface 5588404402c0, with [Window 55884024b110: title: "DayZ"]
[LOG] Searching for matching rules for steam_app_221100 (title: DayZ)
[LOG] [CPrimarySelectionProtocol] CPrimarySelectionProtocol::updateSelection: cannot send selection to a client without a data_device
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] cursorImage request: surface 5588400bafa0
[LOG] CWLSurface 55883f70f6e0 called init()
[LOG] [AQ] drm: Cursor buffer imported into KMS with id 306
[LOG] [AQ] drm: Cursor buffer imported into KMS with id 306
[LOG] [CXDGSurfaceResource] xdg_surface 55883fee4440 requests geometry 0x0 931x1116
[LOG] X11 Activate request for window [Window 55884024b110: title: "DayZ"]
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] [AQ] [libinput] event5  - button state: touch 0 from BUTTON_STATE_NONE    event BUTTON_EVENT_IN_AREA     to BUTTON_STATE_AREA

[LOG] [AQ] [libinput] event5  - tap: touch 0 (TOUCH_BEGIN), tap state TAP_STATE_IDLE → TAP_EVENT_TOUCH → TAP_STATE_TOUCH

[LOG] [AQ] [libinput] event5  - gesture: [1fg] event GESTURE_STATE_NONE → GESTURE_EVENT_FINGER_DETECTED → GESTURE_STATE_UNKNOWN

[LOG] [AQ] [libinput] event5  - gesture: [1fg] state GESTURE_STATE_NONE → GESTURE_STATE_UNKNOWN

[LOG] [AQ] [libinput] event5  - gesture: [1fg] event GESTURE_STATE_UNKNOWN → GESTURE_EVENT_HOLD_TIMEOUT → GESTURE_STATE_HOLD

<<< THE ISSUE HAPPENED AROUND HERE, BUT NO LOGS WERE GENERATED >>>

[LOG] [AQ] [libinput] event5  - tap: touch 0 (TOUCH_UPDATE), tap state TAP_STATE_TOUCH → TAP_EVENT_MOTION → TAP_STATE_DEAD

[LOG] [AQ] [libinput] event5  - gesture: [1fg] event GESTURE_STATE_HOLD → GESTURE_EVENT_CANCEL → GESTURE_STATE_NONE

[LOG] [AQ] [libinput] event5  - gesture: [1fg] event GESTURE_STATE_HOLD → GESTURE_EVENT_POINTER_MOTION_START → GESTURE_STATE_POINTER_MOTION

[LOG] [AQ] [libinput] event5  - gesture: [1fg] state GESTURE_STATE_HOLD → GESTURE_STATE_POINTER_MOTION

[LOG] [AQ] [libinput] event5  - button state: touch 0 from BUTTON_STATE_AREA    event BUTTON_EVENT_UP          to BUTTON_STATE_NONE

[LOG] [AQ] [libinput] event5  - gesture: [1fg] event GESTURE_STATE_POINTER_MOTION → GESTURE_EVENT_END → GESTURE_STATE_NONE

[LOG] [AQ] [libinput] event5  - tap: touch 0 (TOUCH_END), tap state TAP_STATE_DEAD → TAP_EVENT_RELEASE → TAP_STATE_IDLE

[LOG] Keybind triggered, calling dispatcher (8, , 65289, workspace)
[LOG] Searching for matching rules for steam_app_221100 (title: DayZ)
[LOG] [CWLDataDeviceProtocol] New wayland offer 558840348b70 for data source 558840420bc0
[LOG] [CWLDataOfferResource]  | offer 558840348b70 supports mime text/plain;charset=utf-8
[LOG] [CWLDataOfferResource]  | offer 558840348b70 supports mime text/plain
[LOG] [CWLDataOfferResource]  | offer 558840348b70 supports mime TEXT
[LOG] [CWLDataOfferResource]  | offer 558840348b70 supports mime STRING
[LOG] [CWLDataOfferResource]  | offer 558840348b70 supports mime UTF8_STRING
[LOG] Set keyboard focus to surface 55884030f4f0, with [Window 5588405c92a0: title: "foot"]
[LOG] Searching for matching rules for foot (title: foot)
[LOG] [CPrimarySelectionProtocol] CPrimarySelectionProtocol::updateSelection: cannot send selection to a client without a data_device
[LOG] Hyprctl: new connection from pid 73963
[LOG] cursorImage request: shape 1 -> default
[LOG] CWLSurface 55883f70f6e0 called destroy()
[LOG] [AQ] drm: Cursor buffer imported into KMS with id 302
[LOG] [CXDGSurfaceResource] xdg_surface 55883fee4440 requests geometry 0x0 931x1116
[LOG] [CXDGSurfaceResource] xdg_surface 55883fee4440 requests geometry 0x0 931x1116
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963
[LOG] Hyprctl: new connection from pid 73963

Having said so, I have some extra details:

• I was able to reproduce the issue without an external monitor connected. I did not notice before because with the built-in laptop monitor the issue does not happen as often and FPS drops are less dramatic (I originally did not even notice them while gaming).
• I suspect the issue is related to the resolution of the monitor (the external monitor, with an external resolution, causes the GPU to work more, and causes the issue to happen more often). If I increase the quality of the game in the built-in monitor, then the issue becomes more obvious.
• The strange thing is that when increasing the resolution/quality of the game, in addition to the resulting lower FPS (which are expected) I still find this "additional massive drop in FPS once every X seconds".
• The Hyperland monitor still spikes up from ~10% to ~20% (or more) when the issue happens.

Maybe this is not related to Hyprland at all? Maybe this is a Intel driver issue? (...but I tried with two different drivers, i915 and xe, and the issue did not disappeared)

I'll make some more tests (I will also try with the same laptop/kernel but, for example, with gnome, and see what happens).

Sorry for all this mess.

[greenfoo]

I changed the original title of this discussion from Performance hit every X seconds when connecting an external monitor to Performance hit every X seconds when playing games on high resolution/quality to reflect the latest findings.

[greenfoo]

I was able to reproduce the same issue (with the same laptop, same kernel, same everything) on KDE, thus this does not seem to be a Hyprland issue.

HOWEVER there is something to notice: in the KDE case, while FPS drop in the same way, when that happens the kwin_wayland process does not start using more CPU (unlike Hyprland, which spikes from 10% to 20% CPU for ~15 seconds).

Feel free to close / delete this discussion and sorry for this false positive.