TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: 2x Arc A770 or 1x Radeon 7900 XTX for llama.cpp

5 点作者 danielEM2 个月前
Can&#x27;t find &quot;apple to apple&quot; comparison on performance on QWQ 32b (4bit), can anyone help me with decision on which solution to pick?<p>From what I dig so far it looks like dual Arc A770 is supported by llama.cpp. And saw some reports that llama.cpp on top of IPEX-LLM is fastest way for inference on intel card.<p>On the other end there is more expensive 7900 XTX on which AMD claims (Jan &#x27;25) that inference is faster than on 4090.<p>So - what is the state of the art as of today, how does one compare to another (apple to apple)? What is tokens&#x2F;s diff?

2 条评论

runjake2 个月前
I don&#x27;t know but you&#x27;ll probably find a better answer here:<p><a href="https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;LocalLLaMA&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;LocalLLaMA&#x2F;</a><p>Using the search gave me a bunch of threads, but here&#x27;s one:<p><a href="https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;LocalLLaMA&#x2F;comments&#x2F;1ip6c9e&#x2F;looking_to_buy_two_arc_a770_16gb_for_llm&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;LocalLLaMA&#x2F;comments&#x2F;1ip6c9e&#x2F;looking...</a>
评论 #43336655 未加载
laweijfmvo2 个月前
How about a 3090?
评论 #43336695 未加载