vSphere 8U3 supports mixing vGPU profile sizes and types! 😄

Hi all, I'm currently playing around with some vGPU setting in vCenter. For a test I have enabled SR-IOV for my GPU in vCenter, but now I can't disable it. Does anybody know how to do that?

Has anyone managed to get a GeForce and Tesla GPU working on the same machine? What configuration changes did you make? When I install the GeForce drivers, I am not able to use the Tesla GPU, and when I install the Tesla drivers, I am not able to use the GeForce GPU.

Today, I finished a simple guide that anyone can follow and set up.

Setting Up and Configuring On-Prem Nvidia License Server for Nvidia AI Enterprise and Virtual GPU

https://vandu.tech/deploying-and-configuring-nvidia-dls-for-ai-enterprise-and-vgpu-step-by-step-guide/

I have an ESX host with two Nvidia L40 GPU. We have configured a 2Q profile but Nvidia-smi shows only 1824 MIB per VM. Also Windows Task Manager is showing 1,8 GB Also, both GPU are shown as 44,9 GB instead of 48 GB. I'm just wondering if this is by design or maybe a bug?

Hello, I am tasked with scheduling and preparing an upgrade task for ESXi, Citrix, and NVIDIA. However, I am unfamiliar with the NVIDIA component. I have 20 ESXi hosts, each with two NVIDIA A16 GPUs. Additionally, I have a Citrix farm with 200-300 VDI VMs (Windows 10) running on MCS (non-persistent VMs). Currently I'm running on NVIDIA version 15.4, I'm targeting the version 16.7. Could you please advise me on the best practices for upgrading the NVIDIA component (host/guest OS driver)? Is it possible to upgrade the NVIDIA guest OS driver on the master Citrix catalog machine, push the update to the Citrix catalog, and then upgrade the ESXi hosts one by one over a period of one week? Alternatively, would it be necessary to perform this operation in a single session over the weekend? Thanks 🙂

Hello, I'm having a Pro/Contra Discussion with a customer about the installation routine. Usually I'm installing the NVIDIA driver BEFORE the VDA. What is your installation routine? 🙂 thanks

NVIDIA vGPU 17.4 Dropped! Adds support for Horizon 2406, Citrix Virtual Apps and Desktops 7 2407, security updates, and bug fixes! https://docs.nvidia.com/vgpu/17.0/whats-new-vgpu/index.html

Is somebody running a four socket machine (4 physical CPUs installed) with two nvidia L40 cards successfully? Customer having issues with the setup. Poor graphics performance…

Does anybody have experience with using Intel Flex 140/170 GPUs for VDI (mostly interested in Horizon)? Thinking they might be a viable budget-friendly (no licensing costs!) NVIDIA alternative...

A good one to understand / keep in mind, especially with the larger NVIDIA GPUs like L40S is sizing and the impact of GPU channels, especially when using smaller vGPU profiles on those bigger GPUs: https://docs.nvidia.com/vgpu/sizing/virtual-pc/latest/best-practices.html#understanding-gpu-channels

Anyone got expirience with optimizing vGPUs for Premiere Pro? GPU memory usage is insanly high and VMs are having issues when limits are reached.. Doesn't seem like it really good at using swap with memory.

A colleague mentioned that it’s also necessary to update the GPUs’ firmware, which makes sense if you think about it. When I update the host firmware on HP servers, I typically use the SPP. However, I haven’t been able to find any clear guidance on how to update GPU firmware. Does anyone know the recommended process or have any documentation for that? A quick Google search hasn’t turned up anything conclusive.

NVIDIA vGPU 18 dropped yesterday! Docs here: NVIDIA Virtual GPU Software v18.0 - NVIDIA Docs

Hey Members, I am working on a VDI Project for 750users with Win11 with all Adobe CAD and Graphics based application. On Xenserver, How GPU is allotted for the VMs. Need to do some Sizing. Can anyone assist me here with documents or suggestions

We are working on an initiative to get NVidia Grid fully supported on our hypervisor, but are running into roadblocks trying to get NVidia's attention. We have registered for contact in their partner portal and have reached out to many on LinkedIn, with no results. • Does anyone have a suggestion on who to connect with to get the ball rolling? • Is Grid required when simply using pass-through instead of true vGPU? • We are having difficulties finding the drivers without the Grid enablement to use them with pass-through, any guidance would be appreciated.

Hi all, Quick question, if you use vGPU on ESXi, do you leave the VMWare SVGA 3D device enabled or disable it in Windows?

Did anyone here used AMD Processor + NVidia GPU + XenServer for Citrix VDI Solution - Any performance issue observed ?

I'm brand new to NVIDIA GPUs in XenApp servers so bare with me. I'm running the NVIDIA A16-4A GPU profile on 6 vCPU/48 GB VMs. On average, 10 users or less connect to these VMs. So I'm giving about 10 users 4 GB of video RAM to work with. I'm seeing an average of 80% of the GPU Frame Buffer being used at any given time on these servers. Is that a cause for concern or is 80% about right?

anyone available for hire to consult on nvidia vgpu with vmware and CVAD VDI?

WSL2 for server 2025? Thought this wasn't even supported in Win11 with vGPU

Hello all, I have an exisiting Virtual Apps setup (2411 - on-prem) on which I added an nVidia A40 GPU last week with a 2 GB A profile (for virtual apps). Everything works well and graphics performance has increased, but when we change monitors (dock / undock a laptop => external screen to laptop screen), the application is black with only fragments of the video. Does anyone recognize this issue / behaviour? Thanks

Can someone please explain to me how to get GPU acceleration on a Windows Server 2022 OS without the use of Citrix, simply through RDP or VNC or whatever? I just want to add a vGPU to an existing VM, so a specific application on running on it could benefit from the GPU. I added the vGPU, an 8q profile, nvidia RTX license, ... but how can I let the application use the Hardware GPU?

With that announcement of Intel Battlemages able to be bifurcated to have 2 gpus on a single pcie x16 slot, is there any information on those XE2 cards supporting vgpu? Or is Intel only using that on their Flow architecture.

Hi everyone, does anyone know if there’s any technical benefit in using an NVIDIA L40s over an L40 in a 3D graphics/CAD use case? According to this comparison, the L40s seems more optimized for AI workloads, while the L40 is recommended for 3D/CAD—which matches our scenario. Interestingly, we received an offer where the L40s is actually cheaper than the L40, so the customer is considering using L40s in their new ESXi hosts. Key question: Can we mix L40 and L40s in the same environment (vSphere/NVIDIA GRID licensing) without running into issues with driver compatibility or vGPU profiles?

Hello, What do you use for parameters/policy in Citrix in use with nvidia vGPU to have the best graphics quality ? Thanks

Hello, I have 3 ESX with each a NIVIA L4 graphic card. We deployed 3 Win2025 VM, one on each ESX. In the VM, we installed the latest NVIDIA. GPU is configured in Passthrough mode on the ESX. I screen output of nvidia-smi too. vGPU is not working because in Nvidia panel I have nothing to configure. Should I use a nvidia license server for passthrough ? Even if it's the case, I can not configure license server in nvidia panel. What can I show more to debug that ? Thanks

Show esx GPU conf

And did you do server restart?