ESXi 8.0 U3 Upgrade Fails to Detect NVIDIA H100 PCIe GPU

Products

• VMware vSphere ESXi

Issue/Introduction

After upgrading a host to VMware ESXi 8.0 Update 3, the NVIDIA H100 PCIe (80 GB) GPU is no longer visible or accessible within the hypervisor. The following symptoms are observed:

• The device is missing from the vSphere Client hardware inventory.

• Running the following command via ESXi Shell or SSH returns no output: esxcli hardware pci list | grep -i nvidia

• The GPU is not listed in the vmkernel logs during the boot sequence.

• The GPU firmware and device details are missing from the physical server's Remote Management Console (e.g., iLO, iDRAC, or IPMI) during POST.

 

Environment

• VMware vSphere ESXi 8.X 

Cause

• The physical hardware fails to enumerate the PCI device during the POST process, preventing the ESXi hypervisor from detecting the GPU.

Resolution

Follow these steps to isolate the hardware detection issue:

1. Reboot the Problematic ESXi host.

2. Access the remote console and monitor the POST sequence to verify if the NVIDIA GPU is detected during hardware initialization.

3. Verify the GPU firmware details are actively displayed in the iLO console.

4. If the hardware remains undetected by the baseboard management controller, engage the hardware vendor to review physical hardware logs or investigate potential physical seating or fault issues with the GPU card.

Additional Information

NVIDIA GPU not detected in ESXi host after installing drivers, VIBs