<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>MATRIX · GPU-Unlock</title>
  <link>https://derekwang2002.github.io/ai-intelligence/en/projects/gpu-unlock/</link>
  <description>Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations.</description>
  <language>en</language>
  <lastBuildDate>Mon, 02 Nov 2026 16:56:00 GMT</lastBuildDate>
  <atom:link href="https://derekwang2002.github.io/ai-intelligence/en/feeds/projects/gpu-unlock.xml" rel="self" type="application/rss+xml"/>
  <item>
    <title>GPU-Unlock exposes faster tensor paths on Ampere GeForce cards</title>
    <link>https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261018-02/</link>
    <guid isPermaLink="false">ev-20261018-02-published</guid>
    <pubDate>Sun, 25 Oct 2026 15:57:07 GMT</pubDate>
    <category>New</category>
    <category>WATCH</category>
    <description>&lt;p&gt;GPU-Unlock provides a Linux kernel module that changes Ampere GeForce firmware state before the driver loads, enabling a higher FP16/BF16 tensor-core issue rate. The author reports 14%–36% faster LLM prefill with little decode change. It requires rebooting and modifies an unsupported low-level path.&lt;/p&gt;&lt;p&gt;Recommendation: WATCH&lt;/p&gt;&lt;p&gt;Sources: &lt;a href=&quot;https://github.com/Zagitta/gpu-unlock&quot;&gt;Zagitta/gpu-unlock&lt;/a&gt;&lt;/p&gt;</description>
  </item>
</channel>
</rss>
