Skip to content

SteamOS 3.8.10 + 3.9 recovery images fail to reach GUI on Minisforum HX100G (RX 6650M/Navi 23 dGPU + 780M iGPU); Bazzite boots fine #2525

Description

@faux123

Summary

On a Minisforum HX100G mini PC, both the SteamOS 3.8.10 (stable) and
3.9 recovery images fail to reach the SteamOS GUI / gamescope session. The
kernel boots but the Steam Deck UI never appears.

Bazzite on the identical, unmodified hardware boots to its session normally.
That rules out a hardware fault and points at a SteamOS-specific dual-GPU /
display problem on this Phoenix + Navi 23 combination.

Hardware

Component Detail
System Minisforum HX100G (Neptune series)
CPU/APU AMD Ryzen 7 7840HS "Phoenix" (Zen 4, 8c/16t)
iGPU AMD Radeon 780M (RDNA3)
dGPU AMD Radeon RX 6650M (Navi 23 / RDNA2, 28 CU, 8 GB GDDR6)
Display HDMI, connector HDMI-A-1 on amdgpu PCI 0000:03:00.0 (the RX 6650M dGPU, per dmesg)

This is a hybrid / dual-GPU machine (780M iGPU + RX 6650M dGPU). The active
display connector belongs to the discrete GPU.

Steps to reproduce

  1. Flash the SteamOS 3.8.10 stable recovery image to USB; boot the HX100G from it.
  2. Repeat with the SteamOS 3.9 recovery image.
  3. In both cases the system never reaches the SteamOS GUI.
  4. Flash and boot Bazzite on the same hardware → reaches its session normally.

Observed errors (verbatim from dmesg)

[drm] *ERROR* lttpr_caps phy_repeater_cnt is 0x0, forcing it to 0x80
amdgpu 0000:03:00.0: amdgpu [drm] Failed to setup vendor infoframe on connector HDMI-A-1: -22

Notes:

  • 0000:03:00.0 is the discrete RX 6650M; HDMI-A-1 is the active display
    output, so the kernel is driving the dGPU's HDMI connector.
  • -22 is -EINVAL on HDMI vendor-infoframe setup.
  • The lttpr_caps phy_repeater_cnt ... forcing it to 0x80 line also appears on
    working amdgpu systems and is likely benign noise, not the fatal cause. Full
    journalctl -b 0 will follow to pinpoint the gamescope-session failure.

Kernel cmdline workarounds attempted

The following amdgpu parameters were added at boot; none restored the GUI:

  • amdgpu.dc=0 — disable AMD Display Core (legacy display path)
  • amdgpu.sg_display=0 — disable scatter/gather display (a known blank-screen
    workaround on Phoenix / RDNA3)

Additional parameters were also tried; exact strings and the per-parameter
result will be appended.

Diagnostics to follow

  • journalctl -b 0 --no-pager from a failed SteamOS boot (focus on the
    gamescope-session start).
  • cat /proc/cmdline, dmesg, DRM connector status
    (for f in /sys/class/drm/*/status; do echo "$f: $(cat $f)"; done).
  • Bazzite version + uname -r for kernel-branch comparison.
  • Steam ID (so the corresponding System Report can be pulled), available on request.

Why this is likely a SteamOS issue, not hardware

Same machine, same display cable/port: Bazzite reaches its session, both tested
SteamOS recovery images do not. The difference is the OS image (kernel branch +
gamescope-session multi-GPU handling), not the hardware.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions