Mystery System Instability - Threadripper PRO 9995WX - MCE Hardware Error

Notably, a CPU core was definitely dead. I went ahead and just returned it all to Micro Center. Sorry, I couldn’t have been more helpful.

1 Like

It has been a long time since my last report. The instability continued until I updated the BIOS to F12g and made some additional configuration changes.

  • Motherboard: Gigabyte TRX50 AI TOP
  • Use case: Proxmox server
  • Memory profile: EXPO enabled — EXPO I profile

BIOS configuration:

  1. Platform Power — FIXES THE REBOOT ISSUE

Path:
Settings → Platform Power

  • AC BACK → Always Off
  • ErP → Enabled (S4+S5) (mandatory fix); If it still wakes, use Enabled (S5).
  • Power Loading → Enabled
  • Wake on LAN → Disabled
  • Soft-Off by PWR-BTTN → Instant-Off

This entire block removes every source of accidental wake without touching RAM or CPU performance.

  1. CPU Virtualization Optimizations (Proxmox)

Path:
Tweaker → Advanced CPU Settings

Set:

  • SVM Mode → Enabled
  • SMT Mode → Enabled
  • Global C-state Control → Auto
  • Power Supply Idle Control → Typical Current Idle

These maximize Proxmox VM/LXC performance and eliminate AMD idle bugs.

  1. IOMMU & PCIe Stability Config (for Passthrough)

Path:
Settings → Miscellaneous

  • IOMMU → Enabled
  • PCIe ASPM Mode → Disabled
    Prevents PCIe power-state oscillations that cause wake events.

Path:
Settings → IO Ports

  • Above 4G Decoding → Enabled
  • Re-Size BAR Support → Enabled
  • SR-IOV Support → Enabled
  • Initial Display Output → Slot of main GPU (usually PCIe slot 1)
  1. USB Power Behavior

(Handled automatically by ErP)

  • USB power is cut during S5 → prevents USB wake
  1. Boot Configuration

Path:
Boot

  • Fast Boot → Disabled
  • CSM Support → Disabled
  • Boot Priority → Set Proxmox SSD as #1

Pure UEFI is mandatory for clean GPU passthrough.