BSOD on virtual machines running on AMD processors with stop code 0x4E (PFN_LIST_CORRUPT)

Products

• VMware vSphere ESXi

Issue/Introduction

• Random Windows virtual machines experience Blue Screen of Death (BSOD) events with the stop code 0x4E - PFN_LIST_CORRUPT when running on hosts with AMD processors.
• The issue has been observed on systems with the following configuration:

• • CPU: AMD EPYC 9005 Series (Codename: Turin)

  • F/M/S: 26/2/1

  • Current Microcode: 0x0B00211E

  • BIOS Version: Cisco UCS BIOS Version CISCO_PRODUCT_ID.4.3.5f.0

Environment

• VMware ESXi hosts running on AMD EPYC 9005 Series processors

• Guest OS: Microsoft Windows (various versions)

• Hardware platform: Cisco UCS

• VMware vSphere ESXi 7.x

Cause

• The root cause has been identified as a known AMD CPU microcode issue that can lead to memory corruption under certain conditions.
• This issue manifests inside Windows guests as PFN_LIST_CORRUPT (0x4E) BSODs.

Resolution

The problem is resolved by upgrading to a fixed CPU microcode version.

• Affected microcode: 0x0B00211E

• Fixed (good) microcode: 0x0B002151 (or later)

1. Verify CPU F/M/S and Microcode Version:
   
   Check the CPU family, model, stepping, and current microcode revision on the ESXi host.
   
   vsish -e get /hardware/cpu/cpulist/0 | egrep "Family|Model|Stepping|Current Revision"
   
   Example output:  Family: 0x1a Model: 0x02 Stepping: 0x01 Current Revision: 0x0b00211e
   
   
2. Update CPU Microcode: If the CPU F/M/S is 26/2/1 and the microcode version is older than 0x0B002151, update the microcode by updating the System BIOS.
   
   
   • Upgrade the system BIOS to the latest version provided by the hardware vendor.
     
     
   • BIOS updates typically bundle newer CPU microcode.
     
     
   • This is the preferred and most reliable method.
   
   Note: There may be a delay between AMD releasing microcode fixes and hardware vendors incorporating them into BIOS updates.
   
   
3. Validation: After completing the update:

• • Reboot the ESXi host.

  • Confirm that the microcode version is 0x0B002151 or newer.

  • Monitor Windows VMs for recurrence of BSOD events.