Workspace Weekly: Stop Guessing What Your NVIDIA GPU is Doing
March 20, 2026
It is well known that Login Enterprise integrates with NVIDIA nVector. However, if you have not explored the nVector integration yet, it gives you a way to measure graphical endpoint latency and GPU performance in virtual desktop environments, and it is worth a look.
Today, I am adding something new to that story. When something feels slow in a GPU-backed virtual desktop, the hard part isn’t noticing it; it is understanding why, and now there is a faster way to find out.
Session Metrics in Login Enterprise lets you pull Windows performance counters directly from your digital workspaces during a Continuous Test or Load Test, with CPU and memory built in.
For NVIDIA GPU counters, you have always been able to add them, but it meant opening performance monitoring on the digital workspace, copying counter categories, names, and instance strings by hand, then building each metric definition one at a time in Login Enterprise.
One missed character or accidental whitespace, and the metric silently stops collecting. It is the kind of setup work that is easy to get slightly wrong and frustrating to debug. So, I built a script that takes care of all of that automatically.
Figure 1: NVIDIA GPU session metrics displayed alongside CPU and Memory in a Login Enterprise test result.
What the Script Does
Run it once on your NVIDIA GPU-enabled digital workspace, and it handles everything.
This script discovers the GPU, pulls the exact counter strings from the machine itself, creates the session metric definitions in Login Enterprise via the API, and groups them with the built-in CPU and memory metrics into a single session metric group, ready to assign to your tests.
The five GPU counters it sets up are:
- GPU Usage: Overall GPU utilization across the session
- GPU Memory Usage: How much GPU memory the session is consuming
- Framebuffer Usage: Framebuffer demand, a good indicator of graphical workload intensity
- Video Encoder Usage: Encoding load, relevant for remoting protocols and video-heavy workflows
- Video Decoder Usage: Decoding load, useful when streaming or video playback is part of the workload
The script reads the instance details directly from PerfMon on the digital workspace, so there is no hard-coding or guessing counter strings. It works with NVIDIA GPU-enabled virtual desktops and vGPU environments where the NVIDIA GPU PerfMon counter set is present.
GPU Visibility in Load Tests and Continuous Tests
Imagine you are validating a new GPU SKU before rolling it out to a few hundred knowledge workers. You run a Login Enterprise Load Test and watch GPU Memory Usage climb steadily as session count increases. By the time you hit 80% of your target density, the framebuffer is saturated. You catch it before go-live, adjust the vGPU profile, and retest with confidence.
Or if you are running Continuous Tests overnight on a production VDI pool. GPU Usage looks normal most of the week but spikes every morning around 8:30. You cross-reference with logon times and EUX scores, tracing it back to a background process in the gold image that runs at first login, showing up in session metrics before users ever file a ticket.
Figure 2: Assigning the NVIDIA GPU session metric group to a test in the Login Enterprise interface.
These session metrics pair naturally with the NVIDIA nVector latency measurements available through the broader nVector integration for Login Enterprise.
If you are measuring graphical endpoint latency with nVector, the GPU session metrics add the server-side resource view, accompanying the client-side latency data.
But they stand on their own, too. An NVIDIA GPU-enabled digital workspace is all you need to get started. For more on the nVector integration, see the official documentation here: NVIDIA nVector Integration
And if you want to go further, SSIM-based image quality measurement is coming as another companion to this integration. (More on that in a coming Workspace Weekly!)
For a deeper look at session metrics in Login Enterprise: Managing Session Metrics
For viewing your results in Continuous Tests and Load Tests: Viewing Session Metrics
How to Deploy NVIDIA GPU Session Metrics in Login Enterprise
This script is available as a preview integration from the Login VSI GitHub repo. It is ready to try in your environment, and we would love your feedback.
For requirements, usage, and companion guidance, see the documentation here (refer to the section for Create-NvidiaGpuSessionMetricGroup.ps1)
Download Create-NvidiaGpuSessionMetricGroup.ps1 from the Login VSI GitHub repo.
Point it at your Login Enterprise appliance with your API token and run it once per digital workspace GPU environment, then assign the resulting group to your Continuous Test or Load Test in the Login Enterprise interface.
Your next run includes GPU data alongside everything else Login Enterprise is already collecting. No copy-paste errors, missing metrics, or second-guessing the counter strings.
As always, we appreciate your feedback and use your insights to make Login Enterprise better for you.
Stay tuned for more Workspace Weeklies and join our Slack channel #workspace-weekly to share field stories, tips, and interesting finds.
To see all Workspace Weekly posts, visit the Workspace Weekly hub!
Workspace Weekly
