<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Dataflow on Andrew Hunter — Systems Architect &amp; Builder</title>
    <link>https://andrewphunter.com/tags/dataflow/</link>
    <description>Recent content in Dataflow on Andrew Hunter — Systems Architect &amp; Builder</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Thu, 29 Oct 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://andrewphunter.com/tags/dataflow/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>What Makes an RDU Different from a GPU?</title>
      <link>https://andrewphunter.com/writing/what-makes-an-rdu-different-from-a-gpu/</link>
      <pubDate>Thu, 29 Oct 2026 00:00:00 +0000</pubDate>
      <guid>https://andrewphunter.com/writing/what-makes-an-rdu-different-from-a-gpu/</guid>
      <description>&lt;p&gt;I was preparing for an interview when I started looking into SambaNova&amp;rsquo;s SN40L. It is a Reconfigurable Dataflow Unit, or RDU, and I could follow the claims about inference performance before I could explain to myself what the machine was actually doing differently.&lt;/p&gt;&#xA;&lt;p&gt;The interview got rescheduled. The question stuck around.&lt;/p&gt;&#xA;&lt;p&gt;I&amp;rsquo;ve been building inference engines and working on GPU kernels, so I had somewhere to start. But trying to understand the RDU as another kind of GPU kept getting in the way. I had to back up to field-programmable gate arrays, FPGAs, and work forward from there. The connection that finally helped was the idea of giving different parts of a computation their own places on the chip, with paths between them.&lt;/p&gt;&#xA;&lt;p&gt;Once I could picture that, the datacenter implications started to make sense too. To get there, we need to start with what the GPU is doing when a model gives us its next token.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
