ESXi PSOD - IOMMU Fault detected for (<varies>/nvidia) - Reason: 0x6 (PTE not set to allow Read.)

Products

• VMware vSphere ESXi

Issue/Introduction

ESXi hosts with NVIDIA GPU has a PSOD.

• Possibly after updating/upgrading GPU driver.

"IOMMU Fault detected for (<varies*>/nvidia)  Reason: 0x6 (PTE not set to allow Read.)"

Note: (*) Message has been seen to include "vmgfx1" and "vmgpX". It seems this changes based on GPU hardware.

Environment

ESXi 8.0

ESXi 7.0

Cause

Details by GPU vendor NVIDIA:

• Driver version 16
• Driver version 17
  
  

An IOMMU fault causes a purple screen crash with XID error 32 on hypervisor hosts with GPUs based on the NVIDIA Hopper and NVIDIA Ada Lovelace GPU architectures.

Resolution

Resolved in later builds of NVIDIA vGPU software.

• Review NVIDIA Virtual GPU (vGPU) Software Releases for current status of GPU software versions along with articles linked in "Cause" section of this article. Work with GPU vendor to confirm next best steps. NVIDIA may recommend upgrading to latest version of software (e.g., from 17 to 19.2, etc.).