PulseAugur
中
实时 17:10:45
English(EN) Running a 35B MoE (Qwen3.6-35B-A3B) on 2x GTX 1080 Ti in 2026 — Real Benchmarks, and Does the Second GPU Actually Help?

35B MoE模型在双1080 Ti GPU上运行,并辅以CPU RAM

一位用户成功地在两块8年前的NVIDIA GTX 1080 Ti显卡上运行了Qwen3.6-35B-A3B,这是一个拥有350亿参数的混合专家(MoE)模型。该设置利用CPU RAM来存储模型的大部分权重,只有活跃的专家能够放入合并后的22GB显存中。这种配置大约能达到每秒20个token的吞吐量,表明即使是较旧的硬件,通过适当的量化和内存管理技术,也可以运行稀疏MoE模型。 AI

影响 证明了通过仔细优化,较旧的消费级硬件也可以运行大型MoE模型,这可能降低实验的门槛。

排序理由 用户在旧硬件上运行特定模型的基准测试。[lever_c_demoted from research: ic=1 ai=0.7]

在 dev.to — LLM tag 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

35B MoE模型在双1080 Ti GPU上运行,并辅以CPU RAM

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
用户在旧硬件上运行特定模型的基准测试。[lever_c_demoted from research: ic=1 ai=0.7]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
120 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

  1. dev.to — LLM tag TIER_1 English(EN) · byeongsoo kang ·

    2026年用2块GTX 1080 Ti运行35B MoE (Qwen3.6-35B-A3B) — 真实基准测试,以及第二块GPU真的有帮助吗?

    <h2> TL;DR (Quick Answer) </h2> <p>I actually ran <strong>Qwen3.6-35B-A3B</strong> — a 35B-parameter mixture-of-experts model (only 3B active per token) — on a pair of <strong>8-year-old GTX 1080 Ti</strong> cards (22 GB combined). Real, measured numbers:</p> <ul> <li> <strong>Ge…