<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Gpu on Aditya Kurniawan (Akur)</title><link>https://akurniawan.github.io/tags/gpu/</link><description>Recent content in Gpu on Aditya Kurniawan (Akur)</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 16 Aug 2026 00:00:00 +0700</lastBuildDate><atom:link href="https://akurniawan.github.io/tags/gpu/index.xml" rel="self" type="application/rss+xml"/><item><title>CUDA Performance, Part 1: Device Busy Is Not GPU Busy</title><link>https://akurniawan.github.io/posts/gpu-timeline-gaps/</link><pubDate>Sun, 16 Aug 2026 00:00:00 +0700</pubDate><guid>https://akurniawan.github.io/posts/gpu-timeline-gaps/</guid><description>A program can report 98.8% device utilization while leaving its compute units idle 39% of the time. This is what I learned building a TUI for nsys timelines, and why the second number is the only one that matters.</description></item></channel></rss>