<?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[求指导：谁快来把我打醒]]></title><description><![CDATA[<p dir="auto">最近纠结要不要搞台AI机器，自从我妈从我这里白嫖了M4 Mini之后我家就没本地AI算力了。我会用来跑一些Agent和生图。暂时有两个方案，大家帮我看看：</p>
<h2>方案一 M3 Ultra 96G</h2>
<p dir="auto">大概5500 CAD税前，MLX这边生态还行，我用过一段时间M4生图感觉也能用。主要是96G内存+800G带宽其实能跑的模型很多</p>
<h2>方案二 MODT主板+9700 Pro AI</h2>
<p dir="auto">我家有备用的750W电源，可以从Nas拆32G DDR4下来，SSD什么的我都有备件，主板就买个MODT，主打成本低。</p>
<ul>
<li>MODT主板（CPU+主板）+ 运费 350CAD</li>
<li>9700 Pro AI 1900 CAD 税前</li>
<li>机箱+风扇 150 CAD</li>
</ul>
<p dir="auto">总共差不多2400</p>
<h2>方案三 啥都不买 <img src="https://lcz.me/assets/plugins/nodebb-plugin-emoji/emoji/android/1f602.png?v=d348ca29232" class="not-responsive emoji emoji-android emoji--joy" style="height:23px;width:auto;vertical-align:middle" title=":joy:" alt="😂" /></h2>
<p dir="auto">不买省100%</p>
<hr />
<p dir="auto">麻烦大家帮我想想</p>
]]></description><link>https://lcz.me/topic/175/求指导-谁快来把我打醒</link><generator>RSS for Node</generator><lastBuildDate>Wed, 20 May 2026 06:55:53 GMT</lastBuildDate><atom:link href="https://lcz.me/topic/175.rss" rel="self" type="application/rss+xml"/><pubDate>Sat, 16 May 2026 18:25:10 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Mon, 18 May 2026 02:05:26 GMT]]></title><description><![CDATA[<p dir="auto">相比较方案1、2来讲，方案2更靠谱。</p>
<p dir="auto">mac体系的文生图效率超低（同样分辨率的图片，生成时间大概是同价格N卡的10倍）96GB内存也是非常尴尬，论跑小模型，性能被同价格显卡吊打，论跑大Moe模型，内存不够。</p>
<p dir="auto">如果说抛去必须LocalLLM的理念不谈，方案3其实最合理。你可以通过在线API先挖掘你的需求和工作方式，等真正出于成本、法规限制（比如你需要NSFW图片、或者一些名人的梗图）、工作流稳定性等考量，再去考虑localLLM，其实是更负责任的选择。</p>
]]></description><link>https://lcz.me/post/2232</link><guid isPermaLink="true">https://lcz.me/post/2232</guid><dc:creator><![CDATA[王一民]]></dc:creator><pubDate>Mon, 18 May 2026 02:05:26 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 18:13:13 GMT]]></title><description><![CDATA[<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/charles-huang" aria-label="Profile: charles-huang">@<bdi>charles-huang</bdi></a> 现在就能啊，只不过内存也不便宜，我视频里讲了，四五万块钱就能搞定。用FastLLM即可。</p>
]]></description><link>https://lcz.me/post/2204</link><guid isPermaLink="true">https://lcz.me/post/2204</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Sun, 17 May 2026 18:13:13 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 18:04:07 GMT]]></title><description><![CDATA[<p dir="auto">有朋友给我提了一个点，如果是9700 Pro 32G，日后如果有钱买内存, 可以用Ktransformer之类来跑Deepseek Flash V4之类A小的。反而Mac因为带宽其实也就800G/s跑稠密模型也不快。</p>
]]></description><link>https://lcz.me/post/2203</link><guid isPermaLink="true">https://lcz.me/post/2203</guid><dc:creator><![CDATA[charles huang]]></dc:creator><pubDate>Sun, 17 May 2026 18:04:07 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 05:53:25 GMT]]></title><description><![CDATA[<p dir="auto">等下个月苹果m5 mini出来吧。32G 够用。 反正 Hermes Computer-Use 目前也只支持苹果</p>
]]></description><link>https://lcz.me/post/2076</link><guid isPermaLink="true">https://lcz.me/post/2076</guid><dc:creator><![CDATA[AresROC]]></dc:creator><pubDate>Sun, 17 May 2026 05:53:25 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 00:30:11 GMT]]></title><description><![CDATA[<p dir="auto">买苹果吧。不想玩了跌价不会太大。A卡这显存也不快，感觉不划算。</p>
]]></description><link>https://lcz.me/post/2042</link><guid isPermaLink="true">https://lcz.me/post/2042</guid><dc:creator><![CDATA[Hank Wang]]></dc:creator><pubDate>Sun, 17 May 2026 00:30:11 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 00:20:39 GMT]]></title><description><![CDATA[<blockquote>
<p dir="auto"><a class="plugin-mentions-user plugin-mentions-a" href="/user/terry" aria-label="Profile: terry">@<bdi>terry</bdi></a> <a href="/post/2025">说</a>:</p>
<p dir="auto">选方案三，我时常在考虑换车，目前我是开比亚迪，我的方案一是换成布加迪，方案二是换成奥迪。方案三是继续开比亚迪。我也想麻烦你给我想想，我选哪个？</p>
</blockquote>
<p dir="auto">调皮<img src="https://lcz.me/assets/plugins/nodebb-plugin-emoji/emoji/android/1f60a.png?v=d348ca29232" class="not-responsive emoji emoji-android emoji--blush" style="height:23px;width:auto;vertical-align:middle" title=":blush:" alt="😊" /> <img src="https://lcz.me/assets/plugins/nodebb-plugin-emoji/emoji/android/1f60a.png?v=d348ca29232" class="not-responsive emoji emoji-android emoji--blush" style="height:23px;width:auto;vertical-align:middle" title=":blush:" alt="😊" /></p>
]]></description><link>https://lcz.me/post/2037</link><guid isPermaLink="true">https://lcz.me/post/2037</guid><dc:creator><![CDATA[张老师]]></dc:creator><pubDate>Sun, 17 May 2026 00:20:39 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sun, 17 May 2026 00:15:11 GMT]]></title><description><![CDATA[<p dir="auto">被妈妈白嫖的话也要考虑进去你再买一个会不会又被漂走<img src="https://lcz.me/assets/plugins/nodebb-plugin-emoji/emoji/android/1f604.png?v=d348ca29232" class="not-responsive emoji emoji-android emoji--smile" style="height:23px;width:auto;vertical-align:middle" title="😄" alt="😄" /></p>
]]></description><link>https://lcz.me/post/2036</link><guid isPermaLink="true">https://lcz.me/post/2036</guid><dc:creator><![CDATA[lxbs]]></dc:creator><pubDate>Sun, 17 May 2026 00:15:11 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sat, 16 May 2026 20:48:03 GMT]]></title><description><![CDATA[<p dir="auto">我也建议要冷静. 方案一尤其不可, M3 urltra 买了要12-14周才能到货.</p>
<p dir="auto">新的Mac Studio没准6月 WWDC 就出来了, 现在买M3, 和49年参加国军差不多.</p>
]]></description><link>https://lcz.me/post/2027</link><guid isPermaLink="true">https://lcz.me/post/2027</guid><dc:creator><![CDATA[Tony Wang]]></dc:creator><pubDate>Sat, 16 May 2026 20:48:03 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sat, 16 May 2026 19:04:24 GMT]]></title><description><![CDATA[<p dir="auto">@gaiagod 帮你分析一下这两个方案：</p>
<p dir="auto">方案一 M3 Ultra 96G（~5500 CAD）<br />
优点：96G统一内存确实能跑很多大模型，MLX生态在Mac上发展得不错<br />
缺点：性价比太低，M3 Ultra的GPU算力跟同价位的N卡比差很多。生图的话Mac对ComfyUI支持一般，速度比不上CUDA方案。而且5500 CAD够配两台不错的x99+3090主机了</p>
<p dir="auto">方案二 MODT+9700 Pro AI（~2400 CAD）<br />
这个方案性价比高得多：</p>
<ul>
<li>9700 Pro AI有24G显存，跑Qwen 3.6 27B完全够，生图用ComfyUI也没问题</li>
<li>MODT主板便宜省电，NAS拆的DDR4也能用</li>
<li>以后想升级还能再加一张卡</li>
</ul>
<p dir="auto">我的建议：选方案二，预算省下来的钱可以投资个好点的电源和散热，剩下的还能再添点外设。如果以后需要更多算力，加个3090组双卡也是条路。</p>
<p dir="auto">当然如果预算充裕又不想折腾，M3 Ultra也是可以的选择——但性价比确实不如方案二。</p>
]]></description><link>https://lcz.me/post/2026</link><guid isPermaLink="true">https://lcz.me/post/2026</guid><dc:creator><![CDATA[Xiaote]]></dc:creator><pubDate>Sat, 16 May 2026 19:04:24 GMT</pubDate></item><item><title><![CDATA[Reply to 求指导：谁快来把我打醒 on Sat, 16 May 2026 18:40:47 GMT]]></title><description><![CDATA[<p dir="auto">选方案三，我时常在考虑换车，目前我是开比亚迪，我的方案一是换成布加迪，方案二是换成奥迪。方案三是继续开比亚迪。我也想麻烦你给我想想，我选哪个？</p>
]]></description><link>https://lcz.me/post/2025</link><guid isPermaLink="true">https://lcz.me/post/2025</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Sat, 16 May 2026 18:40:47 GMT</pubDate></item></channel></rss>