<?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[Strata 0.1.31 + 3090Ti+ 3080 + 64GB RAM]]></title><description><![CDATA[<h1>LLM 本地推理 Benchmark</h1>
<h2>Hardware</h2>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th>Item</th>
<th>Configuration</th>
</tr>
</thead>
<tbody>
<tr>
<td>GPU</td>
<td>NVIDIA GeForce RTX 3090 Ti, 23028 MiB NVIDIA GeForce RTX 3080, 10240 MiB</td>
</tr>
<tr>
<td>CPU</td>
<td>Intel(R) Core(TM) Ultra 7 270K Plus</td>
</tr>
<tr>
<td>RAM</td>
<td>60Gi</td>
</tr>
<tr>
<td>OS</td>
<td>Ubuntu 26.04 LTS</td>
</tr>
<tr>
<td>Kernel</td>
<td>7.0.0-38-generic</td>
</tr>
<tr>
<td>Driver</td>
<td>595.91.07</td>
</tr>
<tr>
<td>Backend</td>
<td>llama.cpp</td>
</tr>
</tbody>
</table>
<h2>Benchmark Settings</h2>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th>Parameter</th>
<th style="text-align:right">Value</th>
</tr>
</thead>
<tbody>
<tr>
<td>Tool</td>
<td style="text-align:right">llama-benchy</td>
</tr>
<tr>
<td>PP</td>
<td style="text-align:right">2048</td>
</tr>
<tr>
<td>TG</td>
<td style="text-align:right">128</td>
</tr>
<tr>
<td>Depth</td>
<td style="text-align:right">16K / 32K / 64K / 120K</td>
</tr>
<tr>
<td>Concurrency</td>
<td style="text-align:right">1</td>
</tr>
<tr>
<td>Runs</td>
<td style="text-align:right">3</td>
</tr>
<tr>
<td>Cache</td>
<td style="text-align:right">Disabled</td>
</tr>
<tr>
<td>TG Length</td>
<td style="text-align:right">Exact</td>
</tr>
<tr>
<td>Latency Mode</td>
<td style="text-align:right">generation</td>
</tr>
</tbody>
</table>
<h2>Results</h2>
<table class="table table-bordered table-striped">
<thead>
<tr>
<th style="text-align:left">model</th>
<th style="text-align:right">test</th>
<th style="text-align:right">t/s</th>
<th style="text-align:right">peak t/s</th>
<th style="text-align:right">ttfr (ms)</th>
<th style="text-align:right">est_ppt (ms)</th>
<th style="text-align:right">e2e_ttft (ms)</th>
</tr>
</thead>
<tbody>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">pp2048 @ d16384</td>
<td style="text-align:right">487990.14 ± 4576.01</td>
<td style="text-align:right"></td>
<td style="text-align:right">39.82 ± 0.36</td>
<td style="text-align:right">37.78 ± 0.36</td>
<td style="text-align:right">7600.50 ± 239.32</td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">tg128 @ d16384</td>
<td style="text-align:right">93.71 ± 4.94</td>
<td style="text-align:right">94.33 ± 4.92</td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">pp2048 @ d32768</td>
<td style="text-align:right">549207.46 ± 18264.55</td>
<td style="text-align:right"></td>
<td style="text-align:right">65.51 ± 2.16</td>
<td style="text-align:right">63.47 ± 2.16</td>
<td style="text-align:right">13641.16 ± 93.11</td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">tg128 @ d32768</td>
<td style="text-align:right">104.92 ± 6.12</td>
<td style="text-align:right">105.67 ± 6.18</td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">pp2048 @ d65536</td>
<td style="text-align:right">602404.29 ± 2694.60</td>
<td style="text-align:right"></td>
<td style="text-align:right">114.24 ± 0.50</td>
<td style="text-align:right">112.19 ± 0.50</td>
<td style="text-align:right">25405.68 ± 138.26</td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">tg128 @ d65536</td>
<td style="text-align:right">95.69 ± 5.91</td>
<td style="text-align:right">96.33 ± 5.73</td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">pp2048 @ d120000</td>
<td style="text-align:right">620944.40 ± 4961.17</td>
<td style="text-align:right"></td>
<td style="text-align:right">198.61 ± 1.57</td>
<td style="text-align:right">196.57 ± 1.57</td>
<td style="text-align:right">46767.65 ± 10.61</td>
</tr>
<tr>
<td style="text-align:left">Qwen/Qwen3.8-Flash-Next</td>
<td style="text-align:right">tg128 @ d120000</td>
<td style="text-align:right">99.18 ± 4.63</td>
<td style="text-align:right">99.33 ± 4.50</td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
<td style="text-align:right"></td>
</tr>
</tbody>
</table>
<h2>Notes</h2>
<ul>
<li>PP = Prompt Processing</li>
<li>TG = Token Generation</li>
<li>120K depth is used for the ~128K context test so PP/TG still fit inside a 128K context window.</li>
<li>Results are mean ± standard deviation across 3 runs.</li>
<li>Prefix cache avoidance enabled with <code>--no-cache</code>.</li>
<li>Fixed generation length enabled with <code>--exact-tg</code>.</li>
</ul>
]]></description><link>https://lcz.me/topic/2183</link><generator>RSS for Node</generator><lastBuildDate>Mon, 12 Oct 2026 12:24:20 GMT</lastBuildDate><atom:link href="https://lcz.me/topic/2183.rss" rel="self" type="application/rss+xml"/><pubDate>Thu, 08 Oct 2026 07:09:50 GMT</pubDate><ttl>60</ttl><item><title><![CDATA[Reply to Strata 0.1.31 + 3090Ti+ 3080 + 64GB RAM on Thu, 08 Oct 2026 07:16:26 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/23220">说</a>:</p>
<p dir="auto">我弟，这特么不能翻译成汉语吗？</p>
</blockquote>
<p dir="auto">我没在ubuntu上装中文软件。然后跑的内容全在上面。只能勉强先达成英文得了。Windows版本的数据我就可以上中文了。<br />
主要是，我现在打字用的是小鹤双拼+讯飞输入法。不喜欢搜狗的流氓广告。讯飞好像不支持linux。未来ubuntu用的不多。</p>
]]></description><link>https://lcz.me/post/23227</link><guid isPermaLink="true">https://lcz.me/post/23227</guid><dc:creator><![CDATA[Thanaots]]></dc:creator><pubDate>Thu, 08 Oct 2026 07:16:26 GMT</pubDate></item><item><title><![CDATA[Reply to Strata 0.1.31 + 3090Ti+ 3080 + 64GB RAM on Thu, 08 Oct 2026 07:10:36 GMT]]></title><description><![CDATA[<p dir="auto">我弟，这特么不能翻译成汉语吗？</p>
]]></description><link>https://lcz.me/post/23220</link><guid isPermaLink="true">https://lcz.me/post/23220</guid><dc:creator><![CDATA[terry]]></dc:creator><pubDate>Thu, 08 Oct 2026 07:10:36 GMT</pubDate></item></channel></rss>