[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f25pklxmhtr8g4":3},{"slug":4,"category":5,"publishedAt":6,"titleZh":7,"titleEn":8,"summaryZh":9,"summaryEn":10,"models":11,"hardwares":15,"bodyZh":19,"bodyEn":20,"sourceName":21,"sourceUrl":22,"updatedAt":6},"qengine-three-3090-esxi-report-conditions-2026-11-14","ARENA","2026-11-13T23:10:00.910Z","三张3090跑Flash-Next的新报告：334.86 tok\u002Fs不能省掉虚拟化、补丁和MTP","A three-3090 Flash-Next report: 334.86 tok\u002Fs needs its ESXi, patch and MTP conditions","一份新社区报告给出QEngine 1.3.5.7的三卡结果；限功耗、无P2P、tensor补丁和测试阶段必须随数字保留。未独立复现，不能推广为3090通用速度。","A new community report records QEngine 1.3.5.7 on three 3090s. Power caps, missing P2P, a tensor patch and benchmark phases are essential conditions, not optional footnotes. No independent AraoAI reproduction is claimed.",[12],{"id":13,"name":14},11,"Qwen3.8-Flash-Next",[16],{"id":17,"name":18},116,"3x NVIDIA RTX 3090","2026年11月13日，club-3090新增一份三张RTX 3090运行Qwen3.8-Flash-Next的社区报告。作者给出QEngine 1.3.5.7、INT4加MTP，narrative与code两项decode分别为259.20和334.86 tok\u002Fs。它是一台具体机器的自报，不是本站复现，也不是所有3090的预期速度。\n\n关键条件不能删：机器是EPYC 7R13与ESXi 8.0直通，三卡均限320W，没有P2P，使用gpu-unlock tensor补丁。限功耗与补丁对结果的影响方向不同，不能只挑一个因素解释与别人的差距。作者还给出经主机内存中转的通信路径，这不是原生裸机互联的等价替身。\n\n报告同时展示了稳定性与边界测试，但不能把PASS翻译成“能力没有损失”。原帖未给出完整能力题集结果，并记有一次工具调用失败。维护者评论确认这只是该机器的数据，还指出工具把QEngine decode来源误标为llama.cpp日志，以及服务身份字段显示问号；标签问题与吞吐数值有效性应分别核对。\n\n另一个容易混淆的地方是原帖和附件并非同一轮。附件rig报告于06:29:01 UTC生成，10K阶段给出的prefill均值4209.96、decode均值252.05 tok\u002Fs，来自同阶段三次短输出探针。不能拿这组prefill去配原帖code的334.86，也不能将短探针当长输出稳态速度。站内仅把附件的成对数据送入L0草稿审核。\n\n许可也必须锁版本：已回读测试提交5028c50的LICENSE，确为PolyForm Noncommercial 1.0.0，含非商业限制。仓库后续PR说明QEngine从1.3.5.8转为MIT，这份测试仍是1.3.5.7，不能用新版本许可替旧镜像背书。选型先问数字是在什么条件下成立，再问能不能复现到自己的机器。\n\n来源：https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2011\n原始附件：https:\u002F\u002Fgist.github.com\u002Fdgaglioni\u002F7334f51d7f7f9070e0f302f4e80789e6\n许可变更线索：https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fpull\u002F2017","On November 13, 2026, a club-3090 contributor posted a three-RTX-3090 Qwen3.8-Flash-Next report. QEngine 1.3.5.7 with INT4 and MTP recorded narrative and code decode means of 259.20 and 334.86 tok\u002Fs. These are the contributor's measurements, not an AraoAI reproduction or a general speed expectation for 3090 cards.\n\nThe conditions are essential: an EPYC 7R13 system with ESXi 8.0 passthrough, a 320W cap on every GPU, no P2P and a gpu-unlock tensor patch. Power limiting and the patch act in different directions. Host-staged transfers are not equivalent to native peer access, so a comparison cannot assign the whole difference to one factor.\n\nBoundary and soak checks do not establish preserved task ability. The issue lacks a complete quality pack and mentions a tool-call miss. A maintainer confirms that this is a rig-specific community result and flags two reporting defects: a QEngine decode-source label incorrectly names llama.cpp, and service identity is printed as question marks. Label defects and measurement validity require separate checks.\n\nThe issue and attached rig report are also different runs. The attachment was generated at 06:29:01 UTC. Its 10K phase reports paired means of 4209.96 prefill and 252.05 decode tok\u002Fs across three short-output probes. Do not combine its prefill with the issue's 334.86 code result, or represent the short probe as sustained long-output throughput. Only the attachment's paired data is submitted as an L0 draft for review.\n\nLicensing must be versioned too. The LICENSE at tested commit 5028c50 is PolyForm Noncommercial 1.0.0 and restricts commercial use. A later repository PR says QEngine changed to MIT from 1.3.5.8; this test used 1.3.5.7. Do not project a newer license backward. Ask what makes a number valid before asking whether your machine can reproduce it.\n\nIssue: https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2011\nAttachment: https:\u002F\u002Fgist.github.com\u002Fdgaglioni\u002F7334f51d7f7f9070e0f302f4e80789e6\nLicense-change context: https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fpull\u002F2017","club-3090 社区自报及维护者评论","https:\u002F\u002Fgithub.com\u002Fnoonghunna\u002Fclub-3090\u002Fissues\u002F2011"]