[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f2vm6cu7duaakf":3},{"slug":4,"category":5,"publishedAt":6,"titleZh":7,"titleEn":8,"summaryZh":9,"summaryEn":10,"models":11,"hardwares":12,"bodyZh":13,"bodyEn":14,"sourceName":15,"sourceUrl":16,"updatedAt":17},"arc-pro-b65-sycl-runtime-long-prefill-regression-2026-11-17","HARDWARE","2026-11-16T23:05:58.730Z","Arc Pro B65社区回归报告：同一份源码，Intel运行时升级后长提示词变慢","Arc Pro B65 report isolates a long-prefill regression to Intel runtime packages","单份社区报告固定llama.cpp a956a852与Qwen3-Coder-30B-A3B-Instruct Q4_K_M，发现更换Intel软件栈后32K预填充降低、decode反而上升。尚无独立复现，不能外推其他显卡。","One community report holds llama.cpp a956a852 and Qwen3-Coder-30B-A3B-Instruct Q4_K_M constant while changing Intel packages. Long prefill falls while decode rises. No independent reproduction supports generalizing to other GPUs.",[],[],"11月16日，llama.cpp issue #32399给出一份Arc Pro B65的社区自报：固定源码a956a852、模型和参数，仅更换镜像里的Intel运行时组件，32K提示词处理吞吐下降，而生成速度反而提高。它不是新显卡发布，也不是本站实测；截至采集时只有这份报告，评论区没有独立复现。\n\n条件相对明确：单张Arc Pro B65、xe驱动、Qwen3-Coder-30B-A3B-Instruct Q4_K_M，命令为llama-bench -fa 1 -b 4096 -ub 2048 -p 32768 -n 128 -r 3，并设置GGML_SYCL_ENABLE_XMX_ESIMD=0。新栈包含oneDNN 2026.1.0、IGC 2.42.6、NEO 26.40等组件。作者给新栈的pp32768写的是约872–895 tok\u002Fs、tg128约100；换回IGC 2.34.4与NEO 26.18等旧组件、仍保留oneDNN 2026.1.0后，表中为1079与92。区间和约数均按原文保留，不取中点伪装成精确实测。\n\n这份报告的价值不在于给显卡排座次，而在于控制变量。作者先怀疑源码修改，随后固定源码再换软件包；只回退oneDNN没有恢复原表现。同时，IGC与NEO的版本依赖导致两者不能随意混配，因此现有证据还不能把责任单独归给其中一个组件。\n\n使用上的判断也要分场景：长文档的prefill与持续生成的decode可以向不同方向变化。“升级后更快”若只看生成速度，会漏掉长提示词的等待成本。作者还明确说，其他GPU、稠密模型和Windows均未验证，不能据此要求所有Intel用户回退。\n\n本站暂不把该表写成正式跑分记录：原文缺少明确操作系统版本，部分数值也是近似区间。保留候选和缺项，等待可核验的环境与原始输出，比凑齐一条貌似完整的数据更重要。\n\n来源：https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Fissues\u002F32399","A community issue filed on November 16 reports a software-stack regression on one Intel Arc Pro B65. With llama.cpp source a956a852, model and flags held constant, replacing Intel runtime packages reduced 32K-token prompt processing while improving decode. This is neither a new GPU launch nor an AraoAI measurement. At collection time it was a single report with no independent reproduction in the comments.\n\nThe setup uses one B65, the xe driver, and Qwen3-Coder-30B-A3B-Instruct Q4_K_M. The command is llama-bench -fa 1 -b 4096 -ub 2048 -p 32768 -n 128 -r 3 with GGML_SYCL_ENABLE_XMX_ESIMD=0. The current stack includes oneDNN 2026.1.0, IGC 2.42.6 and NEO 26.40. Its table reports approximately 872–895 tok\u002Fs for pp32768 and approximately 100 for tg128. With older IGC 2.34.4 and NEO 26.18 packages while keeping oneDNN 2026.1.0, the corresponding entries are 1079 and 92. We preserve approximate values and ranges rather than manufacture precise midpoint measurements.\n\nThe useful evidence is the controlled comparison, not a GPU ranking. Reverting the initially suspected source change did not restore performance. Reverting only oneDNN did not do so either. Package dependencies prevent freely mixing the IGC and NEO versions, so the report cannot yet identify one component as the sole cause.\n\nPrefill and decode can move in opposite directions. A claim that an upgrade is faster based only on generation speed misses long-prompt waiting time. The author explicitly leaves other GPUs, dense models and Windows unverified; the report is not a blanket recommendation to downgrade.\n\nAraoAI is holding this as a candidate rather than importing a formal benchmark record. The original text lacks an explicit operating-system version, and several results are approximate ranges. Missing evidence should remain visible instead of being filled by inference.\n\nSource: https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Fissues\u002F32399","llama.cpp社区单份自报","https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Fissues\u002F32399","2026-11-16T23:05:58.731Z"]