<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Strix-Halo on Phil Wheat</title><link>https://philwheat.net/tags/strix-halo/</link><description>Recent content in Strix-Halo on Phil Wheat</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Tue, 13 Oct 2026 08:57:00 -0500</lastBuildDate><atom:link href="https://philwheat.net/tags/strix-halo/index.xml" rel="self" type="application/rss+xml"/><item><title>Strix Halo × Qwen3.8 (pwilkin)</title><link>https://philwheat.net/links/strix-halo-llama-cpp/</link><pubDate>Tue, 13 Oct 2026 08:57:00 -0500</pubDate><guid>https://philwheat.net/links/strix-halo-llama-cpp/</guid><description>&lt;p&gt;A full, reproducible stack for running Qwen3.8-27B on Strix Halo (gfx1151): calibrated IQ4_XS weights, an IQ4_XS DFlash2 drafter, the UMA scheduler ring, and retained-PM4 dispatch on a custom-built ROCm userspace — installed from source into your home directory via one installer command, no system ROCm replacement. Their matched-protocol reproduction claims fastest-on-Radeon-8060S by 13–25% over the Vulkan-derived stacks (~26 t/s decode with ~61% draft acceptance, ~257 t/s prefill), with machine-readable benchmark evidence and a broader community leaderboard table. Directly relevant to the Halo box.&lt;/p&gt;</description></item></channel></rss>