OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
display-monitors medium very-common DISPUTE

Get a 120, 144 or 165 Hz panel out of 60 Hz

omarchyarchendeavouroscachyosmanjarohyprlandwaylandnvidiadesktop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

Re-audited on this Omarchy 4.0.2-1 workstation (hyprland 0.56.2-1, kernel 7.1.9, nvidia-open-dkms 610.57.04-1) against the Hyprland 0.56.2 source, the current Hyprland wiki and the two cited GitHub threads read in full. The problem is real and very common, but the record's diagnosis is wrong in a way that sends the reader to the wrong fix, and three of its commands do not work as written. First, the mechanism. The cause claims Hyprland 0.55.x/0.56.x ignores an explicit mode on NVIDIA. It does not. Read at tag v0.56.2, src/output/Monitor.cpp lines 757 to 940 sort every driver mode by closeness to the requested one, keep the best three, append the exact request as a custom mode when the closest is more than 1 off, try them in reverse and take the first that passes m_state.test(). A request that fails the atomic test silently lands on a neighbouring mode, which is exactly why a bad @144 shows up as 120 or 143.91 and looks ignored. The same file shows preferred is not only the EDID preferred timing either: with no resolution in the rule it also queues the first three driver modes behind it, and there is a final catch-all that will try any mode at all. Second, the cause omits the two commonest real reasons a panel sits at 60, and they are the ones with different answers: a link that cannot carry the mode, and an EDID that does not offer it. This machine demonstrates the first. hyprctl monitors all here reports a 165 Hz Microstep MSI G274QPF on HDMI-A-1 whose availableModes top out at 2560x1440@144.00Hz, with no 165 Hz entry anywhere, because HDMI 2.0b cannot carry it. Current mode reads 2560x1440@144.00101. The record as written would have a reader edit monitors.lua forever chasing a mode the kernel never exposed. The modeless EDID case is real enough that Omarchy ships a detector for it, /usr/share/omarchy/bin/omarchy-hyprland-monitor-modeless, whose own comment says a monitor powered off at boot answers with a partial EDID carrying no video modes. Third, three command defects. cat /sys/module/nvidia_drm/parameters/modeset returns Permission denied unprivileged, confirmed by running it here, because the file is -r-------- root root. sudo mkinitcpio -P rebuilds nothing on Omarchy 4, confirmed here: /etc/mkinitcpio.d/ contains zero files, so the command aborts with no presets found. The rebuild is sudo limine-mkinitcpio. And the danger field tells the reader to run pacman -Syu, which the Omarchy ALPM guard blocks. Fourth, an omission the reader will trip on: highrr sorts by rounded refresh rate first and resolution second, confirmed in the addBest3Modes lambda, so on a panel offering 1080p240 and 1440p165 it silently drops the resolution, and the wiki states the predefined modes cannot be combined. Fifth, the whole NVIDIA modeset step is now largely obsolete: Arch has enabled DRM since nvidia-utils 560.35.03-5, the 610 module default is on, and Omarchy's installer writes the file anyway. On sources, both hyprland-wiki blob URLs 404, confirmed with curl: the content moved to content/configuring/core/monitors/ and content/nvidia/, so both go in sources_remove and are replaced with the canonical pages that return 200. The two kept sources do support the symptom but need reading with care and the record should not lean on either as a mechanism. omarchy pull request 161 is real, it reports exactly this (preferred landing on 60 Hz on a 4090 with a 120 Hz panel, highrr fixing it), and it is CLOSED and never merged, which is confirmed by Omarchy 4 still shipping mode = "preferred" in /usr/share/omarchy/config/hypr/monitors.lua, read on disk here. Hyprland issue 15210 exists but was auto-closed by the github-actions bot with "Users are no longer allowed to open issues themselves" fourteen seconds after it was opened, so it carries no maintainer triage and its own root cause analysis is a user's guess that the source contradicts. Severity medium and frequency very-common both stand and are left alone. NOT exercised: I changed no mode, scale or config on this workstation and ran no hyprctl keyword, reload or dispatch, so the corrected fix is argued from the 0.56.2 source, the wiki and read-only hyprctl monitors output rather than from a mode I set and watched. I also could not observe a REJECTED available mode line here, because every mode this box uses applies cleanly and the compositor log had rotated past session start. The commands in steps 1 to 3 were each run read-only on this machine and produce the output described.

The cause above was rewritten on 2026-09-13 to match this note; the fix was corrected by the audit itself.
RISK

Only the NVIDIA branch touches anything risky. Rebuilding the initramfs while a DKMS driver build is half finished can leave an image referencing a module that does not exist, which boots to a black screen, so confirm dkms status reports installed for every kernel first. Direct pacman -Syu is blocked on Omarchy by an ALPM guard and a bare pacman -Sy is a partial upgrade, so update with omarchy update and rebuild only after it has finished. An interrupted sudo limine-mkinitcpio leaves an unbootable UKI, so do not power cycle while it runs. Changing a monitor mode itself is safe and reversible: a mode the hardware cannot set is rejected and the compositor falls back to a working one rather than blanking permanently.

Symptom

A 120, 144, 165 or 240 Hz panel runs at 60 Hz, or at some rate below the one the panel is sold at. hyprctl monitors reports the low rate as the current mode. Two different shapes look identical from the chair: either the fast mode is missing from availableModes entirely, or it is listed there and a matching mode in the config still does not take. There is no config error and no message on screen. Motion feels sluggish and mouse movement stutters.

Cause

Four different faults land on the same low refresh rate, and they need different answers.

  1. **The link cannot carry the mode.** HDMI 2.0b carries about 14.4 Gbps of payload, which is not enough for 2560x1440 at 165 Hz, and a DisplayPort link that trains at fewer lanes or a lower bit rate loses its top modes. A marginal cable can train and then fail the modeset. When the link is the limit the fast mode is usually absent from availableModes altogether, because the kernel filters modes the link cannot carry.
  1. **The EDID does not offer the mode.** A monitor powered off at boot answers with a partial EDID carrying no video modes at all, which is the case Omarchy detects with omarchy-hyprland-monitor-modeless. A KVM, a splitter or a long passive adapter can present a reduced EDID the same way.
  1. **Hyprland is asked for a mode it cannot set and silently lands on a neighbour.** On 0.56.2 an explicit mode is not ignored. CMonitor::applyMonitorRule in src/output/Monitor.cpp sorts every driver mode by closeness to the request, keeps the best three, appends the exact request as a custom mode when the closest match is more than 1 off, then tries them in reverse order and takes the first that passes the atomic test. A request for @144 that fails the test therefore lands on 120 or on 143.91, and @165 on a link that tops out at 144 lands on 144, with no config error either way. The rejected attempts are logged at error level as Monitor <name>: REJECTED available mode ....
  1. **No mode is set, so preferred is in force.** With no resolution in the rule, Hyprland tries the EDID preferred timing first and then the first three modes the driver reports. On many panels the preferred timing is the 60 Hz descriptor rather than the fast one. Omarchy 4 ships mode = "preferred" in the catch-all rule of /usr/share/omarchy/config/hypr/monitors.lua, so this is the state out of the box. A request to change that default to highrr was opened as omarchy pull request 161 and closed without merging.

Fix

**1. List every mode the driver actually offers.** This is the command that settles which of the four causes you have, and it needs no root:


hyprctl monitors all -j | jq -r '.[] | "\(.name)  \(.description)", (.availableModes[] | "    \(.)")'

**2. Read that list before editing any config.**

  • **The fast mode is absent.** The link or the EDID is the limit, not Hyprland, and no config line conjures a mode the kernel did not expose. Check the physical path: a 165 Hz 1440p panel needs DisplayPort, because HDMI 2.0b cannot carry that mode at all. Try a shorter or certified cable, try a different port, and power the monitor on before the machine boots.
  • **availableModes is empty, or the monitor sits at 0x0.** That is the modeless EDID case. Omarchy ships a check for it:

omarchy-hyprland-monitor-modeless; echo $?   # 0 means an enabled monitor has no modes

Power cycle the monitor and replug the cable. Hyprland re-reads the EDID on hotplug.

  • **The fast mode is listed.** Go to step 3.

**3. Ask for a mode that is in the list, copied verbatim.** The refresh rate has to match a listed rate, not the number printed on the box. In ~/.config/hypr/monitors.lua:


hl.monitor({ output = "DP-1", mode = "2560x1440@143.91", position = "0x0", scale = 1 })

On Hyprland 0.54 and older, in monitors.conf: monitor=DP-1,2560x1440@143.91,0x0,1.

Then read back what was actually set, because the fallback is silent:


hyprctl -j monitors | jq -r '.[] | "\(.name) \(.width)x\(.height)@\(.refreshRate)"'

If the rate that comes back is not the one you asked for, the mode failed the atomic test rather than being ignored. The compositor log names what it rejected:


grep -E 'REJECTED|using available mode|using custom mode' "$XDG_RUNTIME_DIR"/hypr/*/hyprland.log

A mode that is listed but rejected is almost always bandwidth. Step one rate down the list and see whether that holds.

**4. highrr is a shortcut with a cost.** It picks the highest refresh rate and only then the resolution, so on a panel offering both 1920x1080@240 and 2560x1440@165 it gives you the 1080p mode. The predefined modes cannot be combined, so there is no way to ask for highest resolution and highest rate together. Use it only when you have read the mode list and know the top rate belongs to the resolution you want:


hl.monitor({ output = "", mode = "highrr", position = "auto", scale = 1 })

**5. NVIDIA: DRM mode setting is rarely the fault any more.** Arch has enabled it since nvidia-utils 560.35.03-5, the 610 series module defaults to on in the driver itself, and Omarchy's installer writes /etc/modprobe.d/nvidia.conf regardless. The parameter file is mode 0400, so an unprivileged cat returns Permission denied rather than an answer:


sudo cat /sys/module/nvidia_drm/parameters/modeset   # Y

Only if that prints N, write the file and rebuild the initramfs:


sudo tee /etc/modprobe.d/nvidia.conf >/dev/null <<'EOF'
options nvidia_drm modeset=1
EOF

The rebuild command is not the same on both systems:


sudo limine-mkinitcpio   # Omarchy 4, which boots a UKI through Limine
sudo mkinitcpio -P       # plain Arch, which has presets in /etc/mkinitcpio.d/

On Omarchy 4, mkinitcpio -P aborts with no presets found because /etc/mkinitcpio.d/ is empty, so it rebuilds nothing and the reboot changes nothing.

Verify


hyprctl -j monitors | jq -r '.[] | "\(.name) \(.width)x\(.height)@\(.refreshRate)"'

The rate you asked for has to come back as the *current* mode. Finding it in availableModes proves only that the driver offers it, which was true before the change as well.

Sources

monitor-stuck-at-60hz-high-refresh