Pin a 4K monitor to 120 Hz when its 144 Hz DSC mode commits but shows nothing
Read drm/amd 5033 in full: Krackan Point (Radeon 840M) ThinkPad X13 Gen 6, Corsair 32UHD144 over USB-C SST, 4K@144 committed with dsc: 1 at 1265750 kHz on HBR3 x4 and the panel stays dark, 4K@120 works on the same cable, Windows runs 144 Hz, and through a Lenovo USB4 Dock 5000 the switch hits the WARNING in update_dpia_stream_allocation_table with a USB topology reset. The issue is still open. The bandwidth arithmetic matches the issue's pixel clock and its dcdebugmask=4 test, which prunes the mode without DSC. Confirmed in amdgpu_dm_debugfs.c on torvalds master that connector debugfs exposes dsc_clock_en, link_settings and dp_dsc_fec_support. Confirmed on this workstation that /usr/share/omarchy/config/hypr/monitors.lua ships the catch-all hl.monitor({ output = "", mode = "preferred", ... }), and the Hyprland wiki (hyprwm/hyprland-wiki content/configuring/core/monitors) documents desc: matching, highrr and the Lua hl.monitor fields the fix uses. Not exercised: no affected hardware here, hyprctl eval was not run, and the DSC state at 120 Hz is an inference from the bandwidth (4K@120 fits HBR3 x4 uncompressed), not something the issue measured. The issue was reported under GNOME and Xorg on Ubuntu, so Hyprland's reporting of the dead mode is inferred from the kernel accepting the commit.
Symptom
4K 144 Hz monitor on USB-C or DisplayPort, AMD laptop. hyprctl monitors says the output is running 3840x2160@144 and there is no config error, but the monitor shows no picture or reports no signal. Setting 3840x2160@120 brings the picture straight back on the same cable and port. Windows runs 144 Hz on the same hardware. Through some USB4 docks, switching to 144 Hz also resets every USB device on the dock.
Cause
4K at 144 Hz with 8-bit colour needs about 1265.75 MHz × 24 bpp, roughly 30.4 Gbps, while a 4-lane DisplayPort 1.4 HBR3 link carries about 25.9 Gbps of payload, so the mode only fits with Display Stream Compression. On the reported Krackan Point machine (drm/amd 5033, open) amdgpu selects DSC, trains HBR3 on 4 lanes, commits the stream and reports success, yet the monitor displays nothing. Through a Lenovo USB4 dock the same switch hits a kernel WARNING in update_dpia_stream_allocation_table and a USB topology reset. Because the kernel accepts the commit, Hyprland sees a working mode and never falls back to a neighbouring rate. Omarchy's default catch-all rule uses mode = "preferred", so a monitor whose EDID prefers 144 Hz lands in this state on first plug.
Fix
**1. Get the picture back.** From the laptop panel, or over ssh (hyprctl -i 0 picks the running instance there):
hyprctl monitors all -j | jq -r '.[] | "\(.name) \(.width)x\(.height)@\(.refreshRate) desc:\(.description)"'
hyprctl eval 'hl.monitor({ output = "DP-2", mode = "3840x2160@120", position = "auto", scale = 1.5 })'
**2. Confirm DSC was the difference** (AMD, root needed):
sudo sh -c 'for f in /sys/kernel/debug/dri/*/DP-*/dsc_clock_en; do echo "$f: $(cat $f)"; done'
A 1 on the connector at 144 Hz and 0 at 120 Hz matches this record.
**3. Pin the working rate.** In ~/.config/hypr/monitors.lua, add a rule for this monitor below the catch-all. Matching on desc: survives a port change. Copy the rate verbatim from availableModes:
hl.monitor({ output = "desc:<description from step 1>", mode = "3840x2160@120", position = "auto", scale = 1.5 })
On Hyprland 0.54 and older (hyprlang): monitor = desc:<description>, 3840x2160@120, auto, 1.5.
**4. Things that do not help** on the reported hardware: changing cables, because 120 Hz already works on the same cable, and asking Hyprland for highrr, which picks the broken 144 Hz mode. Watch drm/amd 5033 for a kernel fix before removing the pin.
Verify
hyprctl monitors shows the output at the pinned 120 Hz rate after a reboot and after replugging, and the monitor shows a picture.