I recently added a 970 evo plus NVMe to my system. I created a new Win10 VM and have been experiencing slow network read speeds.
System specs in signature.
VM specs (VM version 13):
6 vCPU's (was on 8, no difference)
8 GB memory
HD 1: 170GB on NVMe datastore using NVMe controller (I've tried SAS, no difference)
NVMe controller (Tried SCSI SAS controller, see above)
VMXNET3 (Tried E1000 and E1000e but performed worse)
-When Running CrystalDiskMark I 1700MB/s+ seq read/writes and 600MB/s read/write 4KiB Q8T8.
-When I try to download a file using SMB from a NAS VM on the same host it starts at 200MB/s but within a couple seconds drops to 20MB/s and bounces from 50MB/s down to 2MB/s.
-Trying E1000 and E1000e performs slightly less (maybe?).
-Doing the same transfer from another win10 VM (VM version 13) I have on a SATA SSD drive on the same host results in consistent 125MB/s using E1000 virtual network adapter.
-iperf3 shows consistent 5GB/s and iperf -R shows 6GB/s.
-HPP enabled on host and confirmed it grabbed the 970 EVO. No difference.
<code>
t10.NVMe____Samsung_SSD_970_EVO_Plus_500GB__________3A38B09153382500
Display Name: Local NVMe Disk (t10.NVMe____Samsung_SSD_970_EVO_Plus_500GB__________3A38B09153382500)
Has Settable Display Name: false
Size: 476940
Device Type: Direct-Access
Multipath Plugin: HPP
</code>
Not sure exactly what else to try. Writing to the drive without network is fine. Transfers on the network without the drives seems to work fine. But using both at the same time slows down bigtime...
SuperMicro X9DRH-7F
Dual E5-2667 v2
88gb DDR3 ECC
ESXI 6.7 u3
