Workspace Weekly: Latency Tells You How Fast. NVIDIA SSIM Tells You How Good.
April 9, 2026
Every VDI team running GPU-accelerated desktops measures latency. Fewer measure what the user sees.
Latency tells you how fast frames are arriving. It does not indicate whether those frames appeared correct when they arrived. A session can have perfectly acceptable latency and still deliver blurry, compressed, or artifact-ridden images to end users. That gap is exactly where user experience complaints originate — the kind that latency metrics alone can’t explain.
The NVIDIA SSIM preview integration for Login Enterprise closes that gap.
Figure 1: SSIM scores displayed in Login Enterprise Platform Metrics during a Continuous Test.
What SSIM Measures
SSIM stands for Structural Similarity Index. It compares what the target desktop rendered with what arrived on the client screen and produces a score representing perceptual image fidelity. Not a subjective rating. Not a user survey. An objective, repeatable, automation-friendly number.
A score close to 1 means what the user sees is a faithful representation of what the desktop produced. A lower score means something in the remoting stack is degrading the image, whether that is codec compression, network conditions, GPU load, or protocol tuning.
Together with nVector latency, you now have two signals instead of one. Latency tells you how fast. SSIM tells you how good. That is the complete picture of GPU-accelerated session quality.
What You Can Do with It
Imagine you are comparing two codec configurations for a graphics-intensive workload. Latency looks similar between them, but SSIM tells a different story: one configuration is quietly compressing the image in ways your users will notice, especially during motion or rendering. You have the data to make the right call before it reaches production.
Or imagine you are running a load test to validate a new GPU profile. As the session count climbs, SSIM scores start dropping before users would think to file a ticket. You catch the degradation threshold, adjust the profile, and retest with confidence.
Or imagine you are doing a cost-per-user analysis across GPU SKUs. Latency is comparable across configurations, but SSIM shows that one SKU maintains higher image fidelity under load. Now you can quantify the quality difference, not just the speed difference, and justify the decision with data.
How It Works
The SSIM integration combines NVIDIA’s nVector agent with its ssim-tool. During a Login Enterprise test, the integration captures screenshots on both the target desktop and the client, compares them using structural similarity analysis, and uploads the resulting scores to Login Enterprise Platform Metrics. The results appear alongside your latency data and everything else you are already tracking.
It runs automatically as part of your existing Login Enterprise continuous or load tests, no manual screenshot comparison, no subjective review.
For the full nVector integration, including latency measurement, see the nVector integration documentation.
Availability & Getting Started
This is a preview integration available today. It is fully functional, publicly available, and ready to use. The scripts are available in the nVector GitHub repo.
For requirements, setup, and companion guidance, check out the nVector integration guide.
It also serves as the foundation for what’s coming: SSIM as a first-class, built-in feature in a future Login Enterprise release, with heatmaps, threshold alerting, and native UI visibility.
If you’re testing GPU-accelerated VDI environments and only looking at latency, you’re only seeing half of the picture.
As always, we appreciate your feedback and use your insights to make Login Enterprise better.
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
