Enable NVIDIA GRID vGPU for Instant-Clone Pools

You can configure NVIDIA GRID vGPU in ESXi hosts and in the golden image in vSphere Client.

The ESXi host assigns GPU hardware resources to virtual machines on a first-come, first-served basis as virtual machines are created. By default, the ESXi host assigns virtual machines to the physical GPU with the fewest virtual machines already assigned. In the vSphere Web Client UI, this mode is labeled best performance mode. (For example, using the vSphere Web Client, you see this mode's settings when you open a host's Edit Host Graphics Settings window and select Shared Direct feature and Spread VMs across GPUs (best performance) feature.)

If you would rather have the ESXi host assign virtual machines to the same physical GPU until the maximum number of virtual machines is reached before placing virtual machines on the next physical GPU, you can use what the vSphere Client UI labels as GPU consolidation mode. In this use case, you configure this mode on each of the vCenter Server's ESXi hosts that have vGPU installed. (For example, using the vSphere Web Client, open a host's Edit Host Graphics Settings window and select Shared Direct feature and Group VMs on GPU until full (GPU consolidation) feature.)

If you are only using a single vGPU profile per vSphere cluster, set the GPU assignment policy for all GPU hosts within the cluster to the best performance mode's settings in order to maximize performance. In this case, you can also have instant-clone pools and full-clone pools that use the same vGPU profile in the same vSphere cluster.

You can have a cluster with some GPU enabled hosts and some non-GPU enabled hosts.

Note: The following considerations apply to the vMotion of vGPU Virtual Machines feature.

• vMotion of vGPU Virtual Machines is supported starting with vSphere 6.7. For details in the vSphere information for how to configure this, look for section "Using vMotion to Migration vGPU Virtual Machines" in the vSphere documentation.
• vSphere Distributed Resource Scheduler (DRS) in vSphere 6.7 Update 1 and later supports initial placement of vGPU VMs without load balancing support.
• DRS in vSphere 6.7 or vSphere 7.0 versions earlier than vSphere 7.0 U3f will not automatically vMotion vGPU VMs when ESXi hosts are placed in maintenance mode. An administrator is required to manually initiate vMotion of vGPU VMs in order to allow ESXi hosts to enter maintenance mode.
• DRS in vSphere 7.0 U3f and later can be configured to allow automatic vMotion when hosts are placed in maintenance mode. See Broadcom KB for instructions. DRS load balancing remains unsupported for vGPU VMs.

NVIDIA GRID vGPU has these potential constraints:

• RDP is not supported.
• The virtual machines must be hardware version 11 or later.
• Horizon 8 does support creating a vGPU instant-clone pool using a cluster with some vGPU enabled hosts and non-vGPU enabled hosts, and will just ignore the non-vGPU enabled hosts when creating the pool. You can not use vMotion to move an instant-clone from a GPU-enabled ESXi host to an ESXi host that does not have GPU hardware configured.

Enable an instant-clone pool to use NVIDIA GRIP vGPU

1. Install NVIDIA GRID vGPU in the physical ESXi hosts.

2. In vCenter Server hardware graphics configuration, select the Host Graphics tab, and in Edit Host Graphics Settings, select Shared Direct.

   The ESXi host uses the NVIDIA GRID card for vGPU.

3. Prepare a golden image with NVIDIA GRID vGPU configured, including selecting the vGPU profile you want to use.

   If you want to persist pciPassthru values to the instant clones, add the pciPassthru configuration parameters to the VM's Advanced parameters.

4. Take a snapshot of the golden image.

5. In Horizon Console, when you create an instant-clone pool, select this golden image and snapshot.

Horizon 8 automatically displays NVIDIA GRID vGPU in the 3D Render field. Horizon 8 also displays the vGPU profile you chose in the golden image. Instant clones inherit the settings configured in the vSphere Client for the golden image.

The vGPU profile can be edited from Horizon Console when creating or pushing images to Instant Clone desktop pools.

Persist NVIDIA Frame Rate Limiter to instant clones

Starting with 2412 release, Horizon 8 will persist the NVIDIA Frame Rate Limiter value set on the snapshot to the instant clones in the pool. You can increase the Frame Rate Limiter value to support graphics-intensive applications.

1. You add the additional pciPassthru configuration parameters to the virtual machine using the vSphere UI's Edit settings > Advanced parameters.

2. Then take the snapshot of the VM and use in the Horizon desktop pool's Schedule Maintenance operations. All of the pciPassthru parameters configured on the snapshot are carried over to the desktop pool's provisioned VMs.

For example, to increase the NVIDIA Frame Rate Limiter from 60 to 120, the required vmx setting is pciPassthru0.cfg.frl_config = 120. From Horizon Agent 2406 and later, the Blast protocol supports 120 FPS.