Open russelltg opened 9 months ago
performance is highly subject to your system, really. On my desktop sway performs slightly worse, on system64's laptop sway performs 3x worse than hl.
Worth noting direct scanout is disabled by default in hyprland.
Similar issue for me on a 4K screen. The Full HD one works flawlessly, while the 4K screen is running at 15-20 FPS.
I have this same issue
I have the same issue running nouveau drivers on the following hardware:
lspci -k | \grep -A 2 -E "(VGA|3D)"
00:02.0 VGA compatible controller: Intel Corporation TigerLake-H GT1 [UHD Graphics] (rev 01)
DeviceName: Onboard IGD
Subsystem: Hewlett-Packard Company TigerLake-H GT1 [UHD Graphics]
--
01:00.0 VGA compatible controller: NVIDIA Corporation GA107GLM [RTX A2000 Mobile] (rev a1)
Subsystem: Hewlett-Packard Company GA107GLM [RTX A2000 Mobile]
Kernel driver in use: nouveau
Here are some screenshots of running glxgears and vkcube:
Notice that I'm getting half the FPS of what the refresh rate is set to, this is not happening in Gnome for example, nor does it happen in sway. Running in 1920x1080 gives me the full FPS, but none of my displays use that resolution.
Is hyprland just too demanding for an iGPU running nouveau?
have you tried disabling nvidia_anti_flicker?
Added:
opengl {
nvidia_anti_flicker = false
}
to my hyprland.conf and ran the benchmarks again, still at the same, low FPS. Also, I'm not using the Nvidia card I think, i', using the iGPU which is Intel wiht nouveau.
Benchmarks are one thing, but I can tell by the mouse cursor, scrolling in vim and scrolling web pages that something is up in Hyprland, that are not present in Gnome, Sway or other DE/WMs.
currently when I have an external monitor 4k plugged in to my notebook (internal 4k) and use my iGPU (i7-10750H) and I disable the internal monitor, so that only the external monitor is used, disable VFR and enable the debug overlay. I get only 30fps constantly no matter what I do tested watching 60fps and 30fps vids.
❯ hyprctl monitors
Monitor DP-3 (ID 1):
3840x2160@60.00000 at 2560x0
description: Samsung Electric Company U32R59x H4ZM600322 (DP-3 via HDMI)
make: Samsung Electric Company
model: U32R59x
serial: H4ZM600322
active workspace: 2 (2)
special workspace: 0 ()
reserved: 0 33 0 0
scale: 1.00
transform: 0
focused: yes
dpmsStatus: 1
vrr: 0
activelyTearing: false
Monitor eDP-1 (ID 0):
2560x1440@60.00000 at 0x0
description: California Institute of Technology 0x1500 (eDP-1)
make: California Institute of Technology
model: 0x1500
serial:
active workspace: 1 (1)
special workspace: 0 ()
reserved: 0 33 0 0
scale: 1.00
transform: 0
focused: no
dpmsStatus: 1
vrr: 0
activelyTearing: false
Hyprland Version
01e5c59d
Bug or Regression?
Bug
Description
Compared to sway (1.8.1), performance is poor even when windows are not currently being resized and no animations are running
How to reproduce
My test setup:
Non-direct scanout test:
VK_ICD_FILENAMES=/usr/share/vulkan/icd.d/intel_icd.x86_64.json mangohud vkcube --present_mode VK_PRESENT_MODE_IMMEDIATE_KHR
running on itSway: 260 FPS, Sway process is using ~10% GPU Hyprland: 150FPS, Hyprland process is using ~50% GPU
Direct scanout test:
vkcube
(same commandline as above) on 4k displaySway: 118FPS, Sway process is using 0% GPU. GPU Power usage is listed as 4W Hyprland: 120FPS, Hyprland process is using 25% GPU. GPU Power usage is listed as 6.4W
I think the direct scanout test ends up memory limited or something, so there is extra GPU headroom for Hyprland's overhead, hence the higher GPU usage and higher power draw but similar FPS from vkcube (unclear on the exact details here, this is a guess)
Let me know if there is any other info I can help with.
Crash reports, logs, images, videos
No response