Pre-flight checklist
Upstream / vanilla kernel check
I tested with a vanilla/upstream kernel and the issue ALSO reproduces there (upstream bug — please report upstream)
Kernel variant
linux-cachyos-bore (BORE, GCC)
System information (cachyos-bugreport.sh)
Manual system information (if cachyos-bugreport.sh is unavailable)
On a laptop with an AMD Radeon RX 7600S (Navi 33) dGPU and Radeon 680M iGPU, the system experiences desktop freezes associated with an AMDGPU GFXOFF operation failing because the SMU does not respond.
The freeze is avoided by disabling the PP_GFXOFF_MASK feature using:
options amdgpu ppfeaturemask=0xffff7fff
However, with this workaround enabled, the system later experiences repeated SMU communication failures when software attempts to query GPU metrics. This causes applications such as btop to hang, even with GPU information disabled in its configuration.
Hardware
Laptop vendor: ASUSTeK
dGPU: AMD Radeon RX 7600S / Navi 33
PCI ID: 1002:7480
PCI address: 03:00.0
iGPU: AMD Radeon 680M / Rembrandt
PCI ID: 1002:1681
PCI address: 77:00.0
Kernel driver: amdgpu
Software
Distribution: CachyOS
Kernel: 7.2.0-1-cachyos-bore
linux-firmware: 1:20260810-2
Desktop: Hyprland / Wayland
Bug description
Problem
Without a custom ppfeaturemask, the desktop can freeze. The relevant AMDGPU error is:
Failed to disable gfxoff! with SMU: No response msg_reg: 29 resp_reg: 0
The workaround below prevents the desktop freeze:
options amdgpu ppfeaturemask=0xffff7fff
This mask disables the 0x8000 GFXOFF feature bit.
Behavior with GFXOFF disabled
With:
options amdgpu ppfeaturemask=0xffff7fff
the current desktop session remains stable and the previous GFXOFF-related freeze has not occurred.
However, the dGPU subsequently begins producing repeated SMU communication failures:
amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
amdgpu 0000:03:00.0: Failed to export SMU metrics table!
These errors repeat whenever GPU metrics or clock information are queried. Eventually:
amdgpu 0000:03:00.0: Failed to get current clock freq!
is also reported.
This causes monitoring applications such as btop to hang or become unusable, including when:
show_gpu_info = "Off"
is configured.
Relevant boot log
The RX 7600S initializes successfully:
amdgpu 0000:03:00.0: smu driver if version = 0x00000035,
smu fw if version = 0x00000040,
smu fw program = 0,
smu fw version = 0x00526000 (82.96.0)
amdgpu 0000:03:00.0: SMU is initialized successfully!
The dGPU and its rings also initialize successfully.
Later, repeated SMU metric requests fail:
amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
amdgpu 0000:03:00.0: Failed to export SMU metrics table!
followed by:
amdgpu 0000:03:00.0: Failed to get current clock freq!
Steps to reproduce
With the workaround
Add:
options amdgpu ppfeaturemask=0xffff7fff
to an AMDGPU modprobe configuration.
Rebuild the initramfs if necessary and reboot.
The GFXOFF-related desktop freeze is avoided.
Start a GPU/system monitoring application such as btop, or otherwise query dGPU metrics.
Observe repeated kernel messages:
SMU: No response msg_reg: 12 resp_reg: 0
Failed to export SMU metrics table!
btop may hang or become unusable, even when GPU information is disabled.
Expected behavior
Expected behavior
The RX 7600S SMU should respond correctly to GFXOFF state transitions and metric/clock queries.
Disabling GFXOFF as a workaround should not cause or expose persistent failures when querying the SMU metrics table.
Actual behavior
Actual behavior
With GFXOFF enabled, an SMU timeout during GFXOFF handling can result in a desktop freeze.
Disabling GFXOFF via ppfeaturemask=0xffff7fff prevents the freeze.
With GFXOFF disabled, repeated SMU metric table requests can time out with msg_reg: 12.
GPU monitoring applications can hang as a result.
Logs / stack traces
[ 8587.152328] RIP: 0033:0x7ff0e1d49a7f
[ 8587.152353] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 00 00 00 00 48 89 44 24 08 48 89 44 24 10 48 8b
[ 8587.152354] RSP: 002b:00007ff0fdc1f028 EFLAGS: 00000246 ORIG_RAX: 0000000000000000
[ 8587.152356] RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 00007ff0e1d49a7f
[ 8587.152357] RDX: 0000000000000001 RSI: 0000000000000565 RDI: 0000000000000023
[ 8587.152357] RBP: 00007ff12accf1f0 R08: 0000000000000031 R09: 0000000000000000
[ 8587.152358] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
[ 8587.152358] R13: 0000000000000005 R14: 0000000000000000 R15: 0000000000000000
[ 8587.152362] ---[ end trace 0000000000000000 ]---
[ 8587.384673] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8587.384611] in param:00000005
[ 8587.384621] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8587.384621] amdgpu 0000:03:00.0: Dumping IP State Completed
[ 8587.384721] amdgpu 0000:03:00.0: [drm] AMDGPU device core dump file has been created
[ 8587.384774] [drm] Check your /sys/class/drm/card1/device/devcoredump/data
[ 8587.384775] amdgpu 0000:03:00.0: ring gfx_0.0.0 timeout, signaled seq=5671512, emitted seq=5671512
[ 8587.384777] amdgpu 0000:03:00.0: Process ps pid 2466 thread qs:cs:0 pid 2566
[ 8587.384781] amdgpu 0000:03:00.0: Starting ring gfx_0.0.0 ring reset
[ 8587.385758] amdgpu 0000:03:00.0: ring gfx_0.0.0 reset succeeded
[ 8587.385772] [drm] device wedged, but no recovery needed
[ 8646.487766] [UFW BLOCK] IN=wlan0 OUT= MAC=33:33:00:01:00:03:ab:59:5f:89:1c:73:08:00:ad SRC=fe80:00:00:00:00:00:00:00:00:00:00:00:00:00:00:00 DST=ff02:00:00:00:00:00:00:00:00:00:00:00:00:00:00:00 LEN=39
[ 8646.485832] [UFW BLOCK] IN=wlan0 OUT= MAC=01:00:5e:00:00:01:fc:a8:59:5f:89:1c:73:08:00 SRC=192.160.11.4 DST=224.0.0.1 LEN=28 PROTO=UDP SPT=1900 DPT=1900
[ 8646.495506] [UFW BLOCK] IN=wlan0 OUT= MAC=01:00:5e:00:00:01:fc:a8:59:5f:89:1c:73:08:00 SRC=192.160.11.4 DST=224.0.0.1 LEN=28 PROTO=UDP SPT=1900 DPT=1900
[ 8647.406218] [UFW BLOCK] IN=wlan0 OUT= MAC=01:00:5e:00:00:01:fc:a8:59:5f:89:1c:73:08:00 SRC=192.160.11.4 DST=224.0.0.1 LEN=28 PROTO=UDP SPT=1900 DPT=1900
[ 8672.597554] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8672.597554] amdgpu 0000:03:00.0: Dumping IP State
[ 8672.599814] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8673.512983] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8673.512990] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8678.180232] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8678.180280] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8682.837712] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8682.837717] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8687.495582] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8687.495589] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8687.495610] amdgpu 0000:03:00.0: Dumping IP State Completed
[ 8687.495656] amdgpu 0000:03:00.0: [drm] AMDGPU device core dump file has been created
[ 8687.495663] [drm] Check your /sys/class/drm/card1/device/devcoredump/data
[ 8687.495665] amdgpu 0000:03:00.0: ring sdma0 timeout, signaled seq=9924, emitted seq=9924
[ 8687.495666] amdgpu 0000:03:00.0: Starting sdma0 ring reset
[ 8687.495682] amdgpu 0000:03:00.0: reset sdma queue (0:0)
[ 8687.496445] [drm] device wedged, but no recovery needed
[ 8707.513741] amdgpu 0000:03:00.0: Dumping IP State
[ 8725.959794] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8725.959801] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8780.619504] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8780.619515] amdgpu 0000:03:00.0: Failed to disable gfxoff!
[ 8795.272495] amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
[ 8795.272504] amdgpu 0000:03:00.0: Failed to disable gfxoff!
Additional system information
Additional notes
This is a hybrid AMD laptop with both an RX 7600S dGPU and Radeon 680M iGPU using the same amdgpu kernel module. Therefore, ppfeaturemask is a module-wide parameter and cannot easily be applied only to PCI device 03:00.0.
The SMU interface version message is also present:
smu driver if version = 0x00000035
smu fw if version = 0x00000040
but the driver immediately reports:
SMU is initialized successfully!
so it is unclear whether the interface-version difference is related to the subsequent GFXOFF or metrics-table communication failures.
Logs
Relevant logs are available from the affected boot and include:
Successful Navi 33 SMU initialization
SMU interface version information
Repeated msg_reg: 12 SMU timeouts
Failed SMU metrics table exports
Failed current clock frequency queries
The affected GPU is:
03:00.0 VGA compatible controller: AMD Navi 33
[Radeon RX 7600/7600 XT/7600M XT/7600S/7700S / PRO W7600]
Pre-flight checklist
linuxkernel.Upstream / vanilla kernel check
I tested with a vanilla/upstream kernel and the issue ALSO reproduces there (upstream bug — please report upstream)
Kernel variant
linux-cachyos-bore (BORE, GCC)
System information (cachyos-bugreport.sh)
Manual system information (if cachyos-bugreport.sh is unavailable)
Bug description
Problem
Without a custom ppfeaturemask, the desktop can freeze. The relevant AMDGPU error is:
Failed to disable gfxoff! with SMU: No response msg_reg: 29 resp_reg: 0
The workaround below prevents the desktop freeze:
options amdgpu ppfeaturemask=0xffff7fff
This mask disables the 0x8000 GFXOFF feature bit.
Behavior with GFXOFF disabled
With:
options amdgpu ppfeaturemask=0xffff7fff
the current desktop session remains stable and the previous GFXOFF-related freeze has not occurred.
However, the dGPU subsequently begins producing repeated SMU communication failures:
amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
amdgpu 0000:03:00.0: Failed to export SMU metrics table!
These errors repeat whenever GPU metrics or clock information are queried. Eventually:
amdgpu 0000:03:00.0: Failed to get current clock freq!
is also reported.
This causes monitoring applications such as btop to hang or become unusable, including when:
show_gpu_info = "Off"
is configured.
Relevant boot log
The RX 7600S initializes successfully:
amdgpu 0000:03:00.0: smu driver if version = 0x00000035,
smu fw if version = 0x00000040,
smu fw program = 0,
smu fw version = 0x00526000 (82.96.0)
amdgpu 0000:03:00.0: SMU is initialized successfully!
The dGPU and its rings also initialize successfully.
Later, repeated SMU metric requests fail:
amdgpu 0000:03:00.0: SMU: No response msg_reg: 12 resp_reg: 0
amdgpu 0000:03:00.0: Failed to export SMU metrics table!
followed by:
amdgpu 0000:03:00.0: Failed to get current clock freq!
Steps to reproduce
With the workaround
Add:
options amdgpu ppfeaturemask=0xffff7fff
to an AMDGPU modprobe configuration.
Rebuild the initramfs if necessary and reboot.
The GFXOFF-related desktop freeze is avoided.
Start a GPU/system monitoring application such as btop, or otherwise query dGPU metrics.
Observe repeated kernel messages:
SMU: No response msg_reg: 12 resp_reg: 0
Failed to export SMU metrics table!
btop may hang or become unusable, even when GPU information is disabled.
Expected behavior
Expected behavior
The RX 7600S SMU should respond correctly to GFXOFF state transitions and metric/clock queries.
Disabling GFXOFF as a workaround should not cause or expose persistent failures when querying the SMU metrics table.
Actual behavior
Actual behavior
With GFXOFF enabled, an SMU timeout during GFXOFF handling can result in a desktop freeze.
Disabling GFXOFF via ppfeaturemask=0xffff7fff prevents the freeze.
With GFXOFF disabled, repeated SMU metric table requests can time out with msg_reg: 12.
GPU monitoring applications can hang as a result.
Logs / stack traces
Additional system information
Additional notes
This is a hybrid AMD laptop with both an RX 7600S dGPU and Radeon 680M iGPU using the same amdgpu kernel module. Therefore, ppfeaturemask is a module-wide parameter and cannot easily be applied only to PCI device 03:00.0.
The SMU interface version message is also present:
smu driver if version = 0x00000035
smu fw if version = 0x00000040
but the driver immediately reports:
SMU is initialized successfully!
so it is unclear whether the interface-version difference is related to the subsequent GFXOFF or metrics-table communication failures.
Logs
Relevant logs are available from the affected boot and include:
Successful Navi 33 SMU initialization
SMU interface version information
Repeated msg_reg: 12 SMU timeouts
Failed SMU metrics table exports
Failed current clock frequency queries
The affected GPU is:
03:00.0 VGA compatible controller: AMD Navi 33
[Radeon RX 7600/7600 XT/7600M XT/7600S/7700S / PRO W7600]