Lately, we see a lot of customers adding GPU cards to their ESXi hosts to support GPU-accelerated VDI’s. In this blog, I will explain how the GPU enabled cluster is added step by step to VMware Horizon DaaS.
Adding additional Desktop Managers for GPU enabled cluster
– Go to VMware Horizon DaaS Service Center.
– Go to Tenants, browse tenants
– Click edit behind the tenant where the Desktop Manager will be added
– Select Appliances and click add appliances
– Fill in the Additional Appliance Config
– Click Create Appliances
Creating GPU Desktop capacity
– Go to VMware Horizon DaaS Service Center
– Go to Configuration, Standard Capacity
– Click on the + sign behind Standard Capacity. Create the Capacity Definition. Remember that the vGPU memory corresponds with the GPU profile that is being used.
– Click on the + sign behind Desktop Models.
– Create one or multiple Desktop Models. vCPU, vRAM, and vGPU will be multiplied by the chosen multiplier.
Adding GPU Desktop Capacity to Desktop Manager
– Go to VMware Horizon DaaS Service Center.
– Go to Tenants, browse tenants
– Click edit behind the tenant where the Desktop Manager will be added
– Go to quotas
– Select the Data Center and select the GPU enabled Desktop Manager.
– Select the correct Desktop Capacity
Assign GPU enabled resources to the Desktop Manager
– Go to VMware Horizon DaaS Service Center.
– Go to Service Grid, Resources.
– Go to Desktop Managers and select the newly created Desktop Manager
– Select Compute Resources
– Select the correct Compute Resources and click assign
– Select the desktop compute Resources and click ok
– Select Tenant and click save (GPU profile is visible)
– Go to Networks and select the correct network and click assign
Supported GPU Profiles
– This information can be retrieved from the Fabric Database:
o Start an SSL connection towards a Tenant appliance
o Start PSQL with the command: psql -Uadmin
o select * from vgpu_profiles;
| Column1 | Column2 | Type | Column4 | GPU Mem | Column6 | Max Resolution |
| 1 | 1 | grid_k100 | 8 | 256MB | 2 | 2560×1600 |
| 2 | 1 | grid_k120q | 8 | 512MB | 2 | 2560×1600 |
| 3 | 1 | grid_k140q | 4 | 1GB | 2 | 2560×1600 |
| 4 | 1 | grid_k160q | 2 | 2GB | 4 | 2560×1600 |
| 5 | 1 | grid_k180q | 1 | 4GB | 4 | 2560×1600 |
| 6 | 2 | grid_m60-0q | 16 | 512MB | 2 | 2560×1600 |
| 7 | 2 | grid_m60-1q | 8 | 1GB | 2 | 2560×1600 |
| 8 | 2 | grid_m60-2q | 4 | 2GB | 4 | 2560×1600 |
| 9 | 2 | grid_m60-4q | 2 | 4GB | 4 | 3840×2160 |
| 10 | 2 | grid_m60-8q | 1 | 8GB | 4 | 3840×2160 |
| 11 | 3 | grid_m10-0q | 16 | 512MB | 2 | 2560×1600 |
| 12 | 3 | grid_m10-1q | 8 | 1GB | 2 | 4096×2160 |
| 13 | 3 | grid_m10-2q | 4 | 2GB | 4 | 4096×2160 |
| 14 | 3 | grid_m10-4q | 2 | 4GB | 4 | 4096×2160 |
| 15 | 3 | grid_m10-8q | 1 | 8GB | 4 | 4096×2160 |
Supported GPU Cards
– This information can be retrieved from the Fabric Database:
o Start an SSL connection towards a Tenant appliance
o Start PSQL with the command: psql -Uadmin
o select * from GPU;
| Column1 | GPU Card | Manufacturer | Column4 |
| 1 | NVIDIAGRID K1 | NVIDIA Corporation | 4 |
| 2 | NVIDIATesla M60 | NVIDIA Corporation | 2 |
| 3 | NVIDIATesla M10 | NVIDIA Corporation | 4 |
One Comment
Leave a comment
Hi Geursen,
could you please add the following line to the commands:
psql -Uadmin fdb -p 6432
Regards,
Florian