Refer to the exhibit. An administrator is configuring a virtual desktop gold master image with P40 NVIDIA vGPU. upon trying to add a vCPU assignment in Prism Element, the administrator noticed that…

Question

Exhibit

Options

• AInstall the NVIDIA GPU drivers on the Gold Master.
• BDisable ECC by running nvdia-smi -e 0.
• CInstall the NVIDIA GPU Manager Driver.
• DCheck if ECC Mode is turned on by running nvdia-smi -q | grep "ECC Mode" -A2

• A
  14% (7)
• B
  78% (38)
• C
  8% (4)

Explanation

The NVIDIA Tesla P40 GPU has ECC (Error Correcting Code) memory enabled by default. When ECC is enabled, it reserves a portion of GPU memory for parity bits, which reduces available framebuffer memory. This can make certain vGPU profiles appear unavailable or unselectable in the hypervisor management UI because the required memory cannot be satisfied. The diagnostic and resolution steps are: (D) First, confirm whether ECC is the cause by running 'nvidia-smi -q | grep "ECC Mode" -A2' to inspect the current ECC status. (B) If ECC is enabled, disable it by running 'nvidia-smi -e 0', which frees up the reserved memory and makes the full set of vGPU profiles available. Options A and C relate to driver installation, which is a prerequisite step but not the cause of this specific 'not selectable' issue when the GPU is already recognized by the system.

No community discussion yet for this question.