AMD GPU Error Remediation - Kernel Downgrade

March 30, 2026 ยท View on GitHub

Problem

Multiple GPU errors on kernel 6.17 with Navi 32:

  • Display Manager Interrupt Delays
  • GPU Scheduler Errors
  • TLB Fence Work Issues
  • Queue Evictions
  • SMU Firmware Mismatch

Solution

  1. Downgraded kernel from 6.17 to 6.11 (LTS stable)
  2. Added GPU stability parameters: amdgpu.gpu_recovery=1 amdgpu.ppfeaturemask=0xffffffff amdgpu.dpm=1
  3. Held kernel 6.17 packages to prevent auto-removal
  4. Set GRUB to show menu with 10s timeout for kernel selection

Originally documented ~Nov 2025.