NVIDIA Nsight Cloud

NVIDIA Nsight Cloud is a suite of cloud-native components that enable NVIDIA Nsight tools to profile and operate in containerized cloud, cluster, data center, and HPC environments. It streamlines the deployment of Nsight tools so you can profile and capture data from the CPU, GPU, network, storage, workload, API, and more, helping you identify and fix performance bottlenecks and optimize workloads.

Quick Links

• Nsight Cloud product page
• Nsight Cloud documentation
• Nsight Cloud forum
• Feedback: nsight-cloud-feedback@nvidia.com

How Nsight Cloud Works

Nsight Cloud addresses the biggest barriers to profiling in cloud and cluster environments: enabling profiling without modifying your containers, controlling active profiling sessions across a fleet, and allowing analysis directly from a cloud environment.

For Kubernetes, the Nsight Operator automates the management of all software components required to profile your workload: the Nsight Injector ensures Nsight tools are available and configured correctly in your pod, the Nsight Coordinator allows you to control the state of profiling sessions across your fleet of systems, and the Nsight Streamer deploys the Nsight Systems user interface in a pod that may be accessed through a web browser.

Quick Start

Nsight Operator on Kubernetes

Install Nsight Operator with Helm:

helm install --wait nsight-operator \
    https://helm.ngc.nvidia.com/nvidia/devtools/charts/nsight-operator-26.2.1.tgz

For the full walkthrough -- labeling workloads, configuring the CLI, starting/stopping profiling sessions, and downloading reports -- see the Quick Start guide.

Nsight Streamer on Kubernetes

Install Nsight Streamer with Helm to access the Nsight Tools UI through a web browser. Replace your-reports-pvc with a PersistentVolumeClaim that contains your .nsys-rep reports:

helm install --wait --set tool=nsys \
  --set volumes[0].name=reports-volume \
  --set volumes[0].persistentVolumeClaim.claimName=your-reports-pvc \
  --set volumeMounts[0].name=reports-volume \
  --set volumeMounts[0].mountPath=/mnt/host/volumes/reports \
  nsight-streamer \
  https://helm.ngc.nvidia.com/nvidia/devtools/charts/nsight-streamer-2026.2.2.tgz

Set tool= to ncu, ngfx, or ndld to stream Nsight Compute, Nsight Graphics, or Nsight Deep Learning Designer instead. See the Nsight Streamer Helm chart page for advanced options (NodePort vs ClusterIP, GPU hardware acceleration, authentication).

Nsight Streamer locally with Docker

If you just want to view existing .nsys-rep files on a workstation, the Nsight Streamer Resources provide a run_nsight_streamer.py helper script that pulls and runs the streamer container directly:

# Linux/Mac
python3 run_nsight_streamer.py -v /home/user/reports nsys

# Windows
python run_nsight_streamer.py -v d:\reports nsys

Replace nsys with ncu, ngfx, or ndld to launch a different Nsight Tool. The script automatically picks free ports, mounts the volume into the container, and (when supported) enables GPU hardware acceleration for video encoding.

What's in This Collection

Helm Charts

• Nsight Operator -- deploys the Nsight Operator and all of its components
• Nsight Streamer -- deploys the Nsight Streamer for browser-based access to the Nsight Tools UI for Kubernetes

Containers

• Nsight Streamer for Nsight Systems -- web UI for Nsight Systems
• Nsight Streamer for Nsight Compute -- web UI for Nsight Compute
• Nsight Streamer for Nsight Graphics -- web UI for Nsight Graphics
• Nsight Streamer for Nsight Deep Learning Designer -- web UI for Nsight Deep Learning Designer

Resources

• Nsight Operator Resources -- CLI scripts and Python dependencies for driving the Operator from outside the cluster
• Nsight Streamer Resources -- helper scripts for running the Nsight Streamer locally with Docker

Key Features

• Automate Profiling Workflows for Kubernetes -- the Nsight Operator manages and deploys Nsight tools across your cluster without requiring changes to your container images.
• Access Remote Diagnostics From Any Browser -- the Nsight Streamer runs the Nsight Tools UI inside a container so you can view profiles via a web browser, keeping your local system free of extra software.
• Coordinate Profiling Across Pods -- the Nsight Coordinator lets users and services toggle profiling across many Kubernetes pods rather than relying on timers or modifying code.
• Analyze Profiled Data as a Service -- run Nsight Systems recipes (statistics, expert systems, regression detection) against collected reports without downloading them locally.

Underlying Nsight Tools

Nsight Cloud is a delivery and orchestration layer for the NVIDIA Nsight developer tools. To learn what each tool does, see its product page:

• Nsight Systems -- system-wide performance analysis: visualize an application's CPU/GPU activity on a unified timeline. This is the tool the Nsight Operator drives by default.
• Nsight Compute -- interactive CUDA kernel profiler with detailed performance metrics and API debugging.
• Nsight Graphics -- frame debugger and profiler with ray-tracing support for Direct3D, Vulkan, OpenGL, OpenVR, and the Oculus SDK.
• Nsight Deep Learning Designer -- IDE for designing and optimizing deep neural networks for high-performance inference.
• Browse all Nsight Developer Tools for the full suite

License Agreements

By pulling and using artifacts from this collection, you accept the terms and conditions of the End User License Agreement.