---
title: "Nsight Compute"
description: "Look inside one GPU kernel with NVIDIA Nsight Compute: replay passes, speed of light, memory workload analysis, occupancy and launch statistics."
canonical_url: "https://fanout.sh/inference-eng/curriculum/12-04"
md_url: "https://fanout.sh/inference-eng/curriculum/12-04.md"
access: "pro-preview"
---

# Nsight Compute

Look inside one GPU kernel with NVIDIA Nsight Compute: replay passes, speed of light, memory workload analysis, occupancy and launch statistics.

- Access: Fanout Pro (public preview)

## Public preview

The two lessons before this one end the same way, with a name. The ranked table the PyTorch profiling lesson built names the operation that owns the device time, and the machine-wide timeline of the Nsight Systems lesson names the kernel, or the gap between kernels, where the time went, and hands everything inside a kernel's own duration to this…

## In this lesson

- Replay and its passes
- What the report should say
- Speed of light
- Memory workload analysis
- Occupancy and launch statistics

## Continue learning

Pro unlocks the complete lesson, video, practice, and code where available.

---
This representation contains public Fanout content only. Protected Pro lessons, account data, billing, checkout, and pricing are not included.

Browse the public content map: https://fanout.sh/sitemap.md