Fix No Signal on a hybrid laptop's dGPU-wired HDMI or USB-C port
Steps 1, 2 and 4 are accurate and match the hyprland-wiki Nvidia page: modeset must read Y, options nvidia_drm modeset=1, and MODULES=(nvidia nvidia_modeset nvidia_uvm nvidia_drm) plus sudo mkinitcpio -P. The page also flags that early KMS can break hibernation resume, which the record omits here. The uwsm file paths are correct: uwsm(1) documents uwsm/env-${compositor} and uwsm/env.d/*. Two problems in step 3. First, AQ_FORCE_LINEAR_BLIT=0 is undocumented and its effect is the opposite of what the framing implies: in aquamarine's src/allocator/GBM.cpp the flag reads !envExplicitlyDisabled("AQ_FORCE_LINEAR_BLIT"), i.e. it is ON by default, and setting it to 0 makes aquamarine try to allocate a LINEAR-modifier buffer for cross-GPU sharing, which the adjacent source comment says 'Nvidia doesn't support', so on the exact hardware this record targets it is more likely to hurt than help. Second, the hyprland-wiki Multi-GPU page explicitly warns against using /dev/dri/cardN paths at all ('Do not use the card1 symlink... It is dynamically assigned at boot and is subject to frequent change'), and gives a udev-symlink recipe instead.
mkinitcpio -P regenerates every initramfs. Run it only after a fully completed pacman -Syu. Doing it mid partial-upgrade, or with a failed nvidia-dkms build, produces an initramfs missing the module and a black screen at boot. Keep the fallback boot entry.
Symptom
On a hybrid Intel/AMD + NVIDIA laptop, the HDMI or USB-C port (which is wired to the dGPU) gives 'No Signal' on the external monitor. Hyprland runs fine on the internal panel. hyprctl monitors never lists the external one. Or the external comes up but is unusably laggy.
Cause
On many laptops the external ports are physically wired to the NVIDIA GPU. If nvidia_drm.modeset is off, or if aquamarine picked the iGPU as the only DRM device, the compositor never drives those connectors. NVIDIA also lacks features needed for good multi-GPU buffer sharing.
Fix
- Confirm DRM modesetting is on. This is the single most common cause:
cat /sys/module/nvidia_drm/parameters/modeset # must be Y
On current Arch, nvidia-utils already ships options nvidia_drm modeset=1 in /usr/lib/modprobe.d/nvidia-utils.conf, so this usually already reads Y. If it prints N:
sudo tee /etc/modprobe.d/nvidia.conf <<'EOF'
options nvidia_drm modeset=1
EOF
sudo mkinitcpio -P
- Load the modules early. In
/etc/mkinitcpio.conf:
MODULES=(nvidia nvidia_modeset nvidia_uvm nvidia_drm)
then sudo mkinitcpio -P and reboot. Note that early KMS can stop resume-from-hibernation working. If that happens, remove these entries again. On Intel hybrids, put i915 first in the list.
- If the external is detected but laggy or broken, tell aquamarine which GPU to use, NVIDIA node first. Do not use raw
/dev/dri/cardNpaths, because those numbers are reassigned at boot. Create a stable symlink instead:
NV_ID=$(lspci -d ::03xx | grep -i nvidia | cut -f1 -d' ')
printf 'KERNEL=="card*", KERNELS=="0000:%s", SUBSYSTEM=="drm", SUBSYSTEMS=="pci", SYMLINK+="dri/nvidia-gpu"\n' "$NV_ID" \
| sudo tee /etc/udev/rules.d/90-nvidia-dev-path.rules
sudo udevadm control --reload
sudo udevadm trigger
ls -l /dev/dri/nvidia-gpu
Then, in ~/.config/uwsm/env-hyprland (this is where the Hyprland wiki says uwsm users should put AQ_* variables, on Arch and Omarchy alike), one export per line:
export AQ_DRM_DEVICES=/dev/dri/nvidia-gpu
Add the iGPU as a second :-separated entry if the internal panel hangs off it. Log out and back in (uwsm stop) for this to take effect. Do not set AQ_FORCE_LINEAR_BLIT=0 here, because it makes aquamarine attempt LINEAR-modifier buffers that NVIDIA does not support.
- If your BIOS offers it, switching from Hybrid to Discrete graphics avoids the problem entirely (remove
optimus-managerfirst, since disabling the service is not enough).
Verify
hyprctl monitors lists the external output with a real mode, and cat /sys/module/nvidia_drm/parameters/modeset returns Y.