Tesla V100-2016 Hyper-V-使用RemoteFX的多个VM

我继承了使用2012 Hyper-V / Fail-over Cluster / Windows VDI构建的系统。
我正在尝试使用Server 2016 Hyper-V（1607）和戴尔R740机箱中的Tesla V100 32GB GPU推出RemoteFX。
许可证门户中的最新网格驱动程序（391.81）安装正常，Hyper-V管理器可以看到该卡并允许它用于RemoteFX。
但是，当我将其添加到VM并尝试启动VM时，我收到错误32788.（“VMName”无法更改状态。操作失败，错误代码为“32788”）。
尝试了新的虚拟机，包括第1代和第2代，以及当前的虚拟机。
如果我在硬件中包含RemoteFX 3D适配器，无论设置如何，VM都会拒绝启动。
它尝试可能1-3分钟，然后在GUI弹出窗口中报告错误32788，它无法启动。
如果我删除虚拟RemoteFX适配器，它会正常启动。
有任何想法吗？



以下为原文

I've inherited a system that is built using 2012 Hyper-V/Fail-over Cluster/Windows VDIs.

I'm trying to roll out RemoteFX using Server 2016 Hyper-V (1607) and a Tesla V100 32GB GPU in a Dell R740 chassis.

Latest grid driver (391.81) from the license portal installs fine and Hyper-V Manager can see the Card and allow it to be used for RemoteFX.

However, when I add it to a VM and try and start the VM I get an error 32788. ("VMName" failed to change state. The operation failed with error code '32788'). Tried new VMs, both Gen 1 and 2, as well as current VMs.

The VM refuses to start if I include a RemoteFX 3D adapter in the hardware, no matter what the settings. It tries for maybe 1-3 minutes, then reports error 32788 in a GUI popup, that it couldn't start. If I remove the virtual RemoteFX adapter it starts fine.

Any ideas?