<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Vllm on hexfusion - Sam Batschelet</title><link>https://hexfusion.io/tags/vllm/</link><description>Recent content in Vllm on hexfusion - Sam Batschelet</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sat, 14 Mar 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://hexfusion.io/tags/vllm/index.xml" rel="self" type="application/rss+xml"/><item><title>Disaggregated Prefill/Decode on Consumer GPUs</title><link>https://hexfusion.io/posts/disaggregated-pd-consumer-gpus/</link><pubDate>Sat, 14 Mar 2026 00:00:00 +0000</pubDate><guid>https://hexfusion.io/posts/disaggregated-pd-consumer-gpus/</guid><description>Running llm-d&amp;rsquo;s disaggregated prefill/decode architecture across an RTX 3060 and a Tesla T4 connected by 25GbE RDMA. What worked, what broke, and what I learned about KV cache transfer at the edge of what consumer hardware can do.</description></item></channel></rss>