<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
<title>Luna — latest tests</title>
<subtitle>Open measurements of a local AI assistant on AMD</subtitle>
<link href="https://vps-46c954c2.vps.ovh.net/feed.xml" rel="self" type="application/atom+xml"/>
<link href="https://vps-46c954c2.vps.ovh.net/" rel="alternate" type="text/html"/>
<id>https://vps-46c954c2.vps.ovh.net/</id>
<updated>2026-10-10T00:00:00Z</updated>
<author><name>g_lejars</name><uri>https://x.com/g_lejars</uri></author>
<entry><title>Generation speed, Vulkan vs ROCm, 1k to 30k context</title><link href="https://vps-46c954c2.vps.ovh.net/#bench"/><id>https://vps-46c954c2.vps.ovh.net/#gen-2026-10-09</id><updated>2026-10-09T00:00:00Z</updated><summary>Vulkan generates +9% (1k) to +23% (30k) faster on an RX 9070 XT.</summary></entry>
<entry><title>Prompt reading, Vulkan vs ROCm, long prompts</title><link href="https://vps-46c954c2.vps.ovh.net/#bench"/><id>https://vps-46c954c2.vps.ovh.net/#pp-2026-10-09</id><updated>2026-10-09T00:00:00Z</updated><summary>ROCm reads long prompts 13–20% faster.</summary></entry>
</feed>
