GPU passthrough / vGPU setup for a gaming check lab workload area – finest practices?

We’re standing up a VMware Cloud Foundation 5.x workload area particularly to host a small inner gaming/emulation check lab – primarily for QA testing sport builds throughout completely different {hardware} profiles (together with some Switch emulation workloads that want Vulkan assist handed by cleanly). Multi-tenant in any other case, however this specific area has some uncommon GPU necessities in comparison with our customary workloads.

Just a few questions for anybody who’s arrange one thing comparable:

• For GPU-intensive gaming/emulation workloads particularly, is direct GPU passthrough (DirectPath I/O) nonetheless the advisable method in VCF 5.x, or has NVIDIA vGPU matured sufficient to deal with Vulkan-heavy workloads with out noticeable overhead in comparison with passthrough?
• Any particular NSX segmentation issues when this workload area wants low-latency GPU scheduling that should not be affected by different tenants’ upkeep home windows?
• Has anybody run into points with SDDC Manager lifecycle administration when a workload area has GPU passthrough configured – particularly round firmware/driver compatibility throughout patching cycles?

We have pretty strict isolation necessities for this area because it’s testing pre-release builds, but additionally want it performant sufficient that graphics-heavy workloads (like resolution-scaled emulation testing) do not bottleneck on virtualization overhead.