Performance Discrepancy with NVIDIA 4000 ADA GPU on vCenter vs. Physical Machine

I’ve been testing the NVIDIA 4000 ADA GPU on a physical tower machine and a Dell 7920 server running ESXi managed by vCenter. Surprisingly, I’m experiencing better throughput results on the Dell tower, which has lower specs than the 7920 (48 cores, 384GB of RAM). I suspect my passthrough parameters are not configured correctly on the 7920 VM.

Here are the passthrough parameters I’m currently using on the 7920 VM:

  • pciPassthru.use64bitMMIO: TRUE
  • pciPassthru.64bitMMIOSizeGB: 64

I’d appreciate any advice or insights from the community on how to optimize these passthrough parameters or if there are other factors I should consider to improve the GPU performance on the 7920 VM.

vCenter Server 8 Essentials

Thank you!

1 Like