Hardware
iMac20,1 (27" 5K, 2020), T2
GPU: Navi 14 [1002:7340], subsystem [106B:0219], rev 47 (Radeon Pro 5500 XT)
Kernel: 7.1.6-1-t2-noble, built 2026-08-03 — i.e. it already includes patch 6001-drm-amd-pm-Fix-boot-problems-in-5300 (dated 2026-07-09), and my device matches all four IDs that patch gates on (0x1002/0x7340/0x106B/0x0219).
Problem
Despite patch 6001 being present and applicable to my exact device, amdgpu still fails intermittently at cold boot with an SMU timeout. Roughly 1 successful boot in 20. On failure the machine hangs before the LUKS prompt; nomodeset avoids it (no acceleration).
Key point: patch 6001 fixes the "SMU: I'm not done with your previous command" variant by hardcoding the allowed-features masks in smu_v11_0_set_allowed_mask. My failure is different: SMU: No response msg_reg: 6 resp_reg: 0 — a complete non-response at EnableAllSmuFeatures (the step right after set_allowed_mask), timing out after ~3 s → -62. So 6001 does not appear to cover this cold-boot case.
Failing boot (7.1.6, ro quiet splash):

initializing kernel modesetting (NAVI14 0x1002:0x7340 0x106B:0x0219 0x47)
use vbios provided pptable
smc_dpm_info table revision(format.content): 4.5
SMU: No response msg_reg: 6 resp_reg: 0
Failed to enable requested dpm features!
Failed to setup smc hw!
hw_init of IP block failed -62
Fatal error during GPU init
probe with driver amdgpu failed with error -62
Successful boot — same kernel, same cmdline (13 min later):

initializing kernel modesetting (NAVI14 0x1002:0x7340 0x106B:0x0219 0x47)
use vbios provided pptable
smc_dpm_info table revision(format.content): 4.5
SMU is initialized successfully!
Display Core v3.2.378 initialized on DCN 2.0
Initialized amdgpu 3.64.0
SMU firmware version matches (0x38/0x38). ppfeaturemask 0xffffbffd / 0xfff7bffd tested, no effect.
Interpretation / question
The identical kernel+cmdline both succeeding and failing points to a boot-time race or stale SMU hardware state left by the Apple firmware, not a mask issue. Could a forced ASIC/SMU reset before hw_init (similar in spirit to patch 7002 for the VI Macs) make cold-boot init deterministic on Navi 14 iMac20,1? Happy to test patches. Full failing + successful dmesg attached.
cc @iurjscsi1101500 (author of patch 6001, same GPU)
Hardware
iMac20,1 (27" 5K, 2020), T2
GPU: Navi 14 [1002:7340], subsystem [106B:0219], rev 47 (Radeon Pro 5500 XT)
Kernel: 7.1.6-1-t2-noble, built 2026-08-03 — i.e. it already includes patch 6001-drm-amd-pm-Fix-boot-problems-in-5300 (dated 2026-07-09), and my device matches all four IDs that patch gates on (0x1002/0x7340/0x106B/0x0219).
Problem
Despite patch 6001 being present and applicable to my exact device, amdgpu still fails intermittently at cold boot with an SMU timeout. Roughly 1 successful boot in 20. On failure the machine hangs before the LUKS prompt; nomodeset avoids it (no acceleration).
Key point: patch 6001 fixes the "SMU: I'm not done with your previous command" variant by hardcoding the allowed-features masks in smu_v11_0_set_allowed_mask. My failure is different: SMU: No response msg_reg: 6 resp_reg: 0 — a complete non-response at EnableAllSmuFeatures (the step right after set_allowed_mask), timing out after ~3 s → -62. So 6001 does not appear to cover this cold-boot case.
Failing boot (7.1.6, ro quiet splash):

initializing kernel modesetting (NAVI14 0x1002:0x7340 0x106B:0x0219 0x47)
use vbios provided pptable
smc_dpm_info table revision(format.content): 4.5
SMU: No response msg_reg: 6 resp_reg: 0
Failed to enable requested dpm features!
Failed to setup smc hw!
hw_init of IP block failed -62
Fatal error during GPU init
probe with driver amdgpu failed with error -62
Successful boot — same kernel, same cmdline (13 min later):

initializing kernel modesetting (NAVI14 0x1002:0x7340 0x106B:0x0219 0x47)
use vbios provided pptable
smc_dpm_info table revision(format.content): 4.5
SMU is initialized successfully!
Display Core v3.2.378 initialized on DCN 2.0
Initialized amdgpu 3.64.0
SMU firmware version matches (0x38/0x38). ppfeaturemask 0xffffbffd / 0xfff7bffd tested, no effect.
Interpretation / question
The identical kernel+cmdline both succeeding and failing points to a boot-time race or stale SMU hardware state left by the Apple firmware, not a mask issue. Could a forced ASIC/SMU reset before hw_init (similar in spirit to patch 7002 for the VI Macs) make cold-boot init deterministic on Navi 14 iMac20,1? Happy to test patches. Full failing + successful dmesg attached.
cc @iurjscsi1101500 (author of patch 6001, same GPU)