<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0"><channel><title><![CDATA[請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm]]></title><description><![CDATA[<p dir="auto">如題，版主yt幾次提到 7900xtx跑 llamacpp 好像都是用 ROCm，想知道 vulkan 是否更好?<br />
用途 : hermess agent 跑 qwen 3.6。原本都用 claude code ，但之後 20美訂閱好像要拿掉 claude code 了想找 opencode hermess agent 或是 open claude code做替代，目前還只做簡單測試，因為開發虛擬機在win11 pro 的 hyper-v上。目前只有用windwos 的 ollama跑 qwen 3.6:27n 跑起來也有 30tk/s。後續想知道有沒有需要換到 linux server。</p>
]]></description><link>https://lcz.me/topic/8/請益-7900xtx-llama.cpp-用-vulkan-還是-rocm</link><generator>RSS for Node</generator><lastBuildDate>Wed, 20 May 2026 07:04:20 GMT</lastBuildDate><atom:link href="https://lcz.me/topic/8.rss" rel="self" type="application/rss+xml"/><pubDate>Sun, 03 May 2026 05:05:32 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Thu, 07 May 2026 15:08:36 GMT]]></title><description><![CDATA[<p dir="auto">謝謝兩位大神回復</p>
]]></description><link>https://lcz.me/post/477</link><guid isPermaLink="true">https://lcz.me/post/477</guid><dc:creator><![CDATA[CHIA AN YANG]]></dc:creator><pubDate>Thu, 07 May 2026 15:08:36 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Thu, 07 May 2026 06:10:52 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/chia-an-yang" aria-label="Profile: CHIA-AN-YANG">@<bdi>CHIA-AN-YANG</bdi></a> 我 7900xt 20G，Windows11 下直接lm studio，rcom，qwen 3.5 或3.6 27b q4km，20k上下文，kv缓存q8，大约23-24 tokens/s</p>
]]></description><link>https://lcz.me/post/414</link><guid isPermaLink="true">https://lcz.me/post/414</guid><dc:creator><![CDATA[jenaflex]]></dc:creator><pubDate>Thu, 07 May 2026 06:10:52 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Wed, 06 May 2026 14:19:22 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/chia-an-yang" aria-label="Profile: CHIA-AN-YANG">@<bdi>CHIA-AN-YANG</bdi></a> 差不多，没问题，ubuntu下也就是25-30</p>
]]></description><link>https://lcz.me/post/346</link><guid isPermaLink="true">https://lcz.me/post/346</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Wed, 06 May 2026 14:19:22 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Wed, 06 May 2026 14:17:16 GMT]]></title><description><![CDATA[<p dir="auto">同主題,想問 我也是7900XTX 24G因為放家裡電腦使用win11, wsl2 ,llama.cpp跑起來26t/s請問這樣速度算還可以嗎?</p>
]]></description><link>https://lcz.me/post/344</link><guid isPermaLink="true">https://lcz.me/post/344</guid><dc:creator><![CDATA[CHIA AN YANG]]></dc:creator><pubDate>Wed, 06 May 2026 14:17:16 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Mon, 04 May 2026 15:27:36 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/%E5%BC%B5%E8%80%80%E6%9D%B1" aria-label="Profile: 張耀東">@<bdi>張耀東</bdi></a></p>
<p dir="auto">TLDR，大部分时候 Rcom好<br />
下面这个博主是个老外软件工程师，做的测试（可以忽略那些 不好用的模型的结果）<br />
<a href="https://kyuz0.github.io/amd-r9700-ai-toolboxes/" rel="nofollow ugc">https://kyuz0.github.io/amd-r9700-ai-toolboxes/</a></p>
]]></description><link>https://lcz.me/post/114</link><guid isPermaLink="true">https://lcz.me/post/114</guid><dc:creator><![CDATA[jenaflex]]></dc:creator><pubDate>Mon, 04 May 2026 15:27:36 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Mon, 04 May 2026 12:58:47 GMT]]></title><description><![CDATA[<p dir="auto">个人感觉，vulkan pp慢 tg快，rocm pp快 tg慢，跑code agnet还是rocm更舒服</p>
]]></description><link>https://lcz.me/post/94</link><guid isPermaLink="true">https://lcz.me/post/94</guid><dc:creator><![CDATA[gggg1234]]></dc:creator><pubDate>Mon, 04 May 2026 12:58:47 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Sun, 03 May 2026 05:40:15 GMT]]></title><description><![CDATA[<p dir="auto">暂时没必要折腾，就等不够了再换</p>
]]></description><link>https://lcz.me/post/31</link><guid isPermaLink="true">https://lcz.me/post/31</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Sun, 03 May 2026 05:40:15 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Sun, 03 May 2026 05:33:02 GMT]]></title><description><![CDATA[<p dir="auto">感謝回答，因為爬reddit 和 github 都是說 Vulkan比 ROCm好。我在 TRAE 剛出就知道 但我習慣用純CLI/TUI介面寫code，拋棄vscode很久了，目前開發環境也是 tmux neovim 直接在 terminal 跑 cli ai agent。打算和版主一樣 過半年再看看 等新技術有人折騰好了再用。</p>
]]></description><link>https://lcz.me/post/30</link><guid isPermaLink="true">https://lcz.me/post/30</guid><dc:creator><![CDATA[張耀東]]></dc:creator><pubDate>Sun, 03 May 2026 05:33:02 GMT</pubDate></item><item><title><![CDATA[Reply to 請益 : 7900xtx llama.cpp 用 vulkan 還是 ROCm on Sun, 03 May 2026 05:16:18 GMT]]></title><description><![CDATA[<p dir="auto">rocm版本，AI会告诉你Vulkan还行，事实上rocm好。你如果能在win下跑到30t/s，那么到linux下也就这个速度。但是Linux下可以折腾下turboquant dflash等技术，vllm之类的。如果只是写代码我认为没必要。另外关于写代码，你可以研究下找个CN代理，下载CN版本的TRAE，估计注册的时候需要接收下短信，它是免费的，体验很好。</p>
]]></description><link>https://lcz.me/post/29</link><guid isPermaLink="true">https://lcz.me/post/29</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Sun, 03 May 2026 05:16:18 GMT</pubDate></item></channel></rss>