Sorry for the delays. The nvme fio test didn't stress the system enough with this build, so I setup two systems with a direct wired, back to back, high performance network link and ran iperf. With the client in pass through, and the server in passthough/nonpassthrough, the data rate was slightly less than half in the nonpassthrough case.
Sorry for the delays. The nvme fio test didn't stress the system enough with this build, so I setup two systems with a direct wired, back to back, high performance network link and ran iperf. With the client in pass through, and the server in passthough/ nonpassthrough, the data rate was slightly less than half in the nonpassthrough case.
This is verified