Occasional laggy animations

[satma0745]

Any animation throughout the system (changing focused window, scroll in browser, some in-app animation) can occasionally lag.

Hardware, versions and logs

Hardware:

• CPU is AMD Ryzen 9700x
• GPU is Nvidia RTX 4070s
• RAM is 32GB dual-channel 6000Mhz
• Display is 3440x1440 160Hz

Versions:

• Endeavour OS (6.17.4-arch2-1) without DE
• Hyprland (v0.51.1, commit 71a1216)
• nvidia driver (580.95.05)

I don't have any crash-reports since nothing crashed. And I also don't what logs could be useful for investigating and fixing such an issue. So I've provided the hyprctl systeminfo -c output: systeminfo.txt.

Problem Description

Basically instead of proper 160Hz (that's the refresh-rate of my monitor) animations occasionally ran in something closer to 30-60Hz. The drop didn't last long - probably something around 500-800ms. I don't think this issue has anything to do with application's framework, because the issue existed in hyprland's ui, firefox, telegram-desktop (Qt app) and other apps.

This hardware is obviously more than capable performance-wise. It also works fine under windows (I'm not trying to downplay hyprland, rather I'm saying that I don't think that the hardware is malfunctioning or broken).

I also don't think that the issue is caused by an Nvidia GPU specifically, because I saw a similar issue from a user with an AMD GPU: #2305.

I noticed that the issue never occurs when animation is already playing. For example, if I go to some website, enable auto-scroll and just watch it, then it runs smooth. But if I stop for a few seconds and try scrolling again, I might (or might not) have a jagged animation.

That's also the case with recording. I tried recording the issue with OBS, but it stopped occurring during the recording (and continued after I stopped recording). After further checks figured out that it depends on the video encoder in use: if I select any of the NVENC options, then animations stop lagging, but if I select a software video encoder, then the issue persists.

Here's the recording from the Telergam Desktop:

2025-10-26.21-26-19.mp4

The actual app here doesn't matter (the issue manifests in other apps and system too), it was just the easier way for me to capture this issue. On the recording it's clearly seen that if I spam messages in a row (when I send "c") there are no issues with animations, but when I wait for a few seconds between messages ("b") the issue manifests.

The recording itself doesn't represent the issue perfectly. When there's no recording, the issue happens less frequently (maybe in 50% of cases if I wait for several seconds and in 10% of cases if I wait for ~1s). Also it looks a little bit differently - what I see with my eyes is closer to this first "a" message that was sent right after switching workspaces. And of course there's a discrepancy in frequency because OBS allowed me to set only 120fps while my display shows 160Hz. Apart from these small differences, the recording is a pretty good representation of how the issue happens and feels.

What fixes didn't work for me

Specifying frequency of the display manually didn't fix the issue for me:

monitor=,3440x1440@160,auto,1.25

Switching to "Performance" preset on the waybar's power-mode out-of-the-box switcher also didn't help. Sorry, I don't know which actual command it performs. It also may seem strange to You that I mentioned it, but it might make sense when You read about what actually helped me.

Specifying the following env vars didn't help. And to be honest, I'm not even sure if I specified them in the correct place (they're in my hyprland.conf, I don't have uwms installed and I launch hyprland from manually by typing Hyprland into tty). Env vars:

env = GBM_BACKEND,nvidia-drm
env = LIBVA_DRIVER_NAME,nvidia
env = __GLX_VENDOR_LIBRARY_NAME,nvidia
env = AQ_DRM_DEVICES, /dev/dri/nvidia-dgpu
env = QT_QPA_PLATFORM,wayland

I went through the following articles:

1. https://wiki.archlinux.org/title/NVIDIA - some of advises were already applied either by EndeavourOS or Hyprland, some were applied by me, some I skipped because they didn't seem relevant/connected.
2. https://wiki.hypr.land/Nvidia
3. https://wiki.hypr.land/Configuring/Multi-GPU - but I specified my nvidia gpu instead of integrated one.

What fix worked for me

Looking at how the issue behaves I got an idea that it might be caused by either GPU or my Display going into some idle mode after a few seconds of not rendering a lot of changes. And the change in behavior during OBS recording made it clear that it's more likely a GPU idling issue.

I used the nvidia-smi app to check what frequencies my GPU had during a NVENC recording (2505 MHz, 10251 MHz) and set these frequencies as new minimum frequencies for the GPU. If these frequencies seem to high for You, they probably are - I didn't test lower values and just used these ones.

@xcpky, @marcoradocchia, noticed that You have had a somewhat similar issue. Maybe this will be a fix for You too.

Why I opened this discussion? Apart from giving a chance for others to discover this issue (btw, noticed that google picks up discussion results less frequently than issues threads for some reason), it also would be nice if this was somehow fixed. If it can't be fixed, maybe it could be at mentioned in the docs (e.g. on the https://wiki.hypr.land/Nvidia page).

[echovl]

I'm also experiencing this exact same issue using a 4060 Ti and nvidia 580 drivers

[gulafaran]

this is more or less an nvidia issue. NVIDIA/open-gpu-kernel-modules#777 NVIDIA/open-gpu-kernel-modules#693 (comment) amongst a few, even tho closed its not completly fixed. as many comments suggest.

"In my opinion both 565 and 570 improved the situation. It's much smoother now. There are still stutters, but not so much."

so yes, setting min gpu clocks higher mitigates it.