{"id":153465,"date":"2025-11-24T00:17:07","date_gmt":"2025-11-23T16:17:07","guid":{"rendered":"https:\/\/www.gocpmall.com\/?p=153465"},"modified":"2025-11-24T00:17:07","modified_gmt":"2025-11-23T16:17:07","slug":"apple-m5-%e6%80%a7%e8%83%bd%e6%9b%9d%e5%85%89%ef%bc%8c%e5%9c%96%e5%83%8f%e7%94%9f%e6%88%90%e5%bf%ab4%e5%80%8d%ef%bc%8c%e6%96%87%e6%9c%ac%e6%8e%a8%e7%90%86%e4%ba%a6%e6%8f%90%e5%8d%87%e8%bf%9130","status":"publish","type":"post","link":"https:\/\/www.gocpmall.com\/?p=153465","title":{"rendered":"Apple M5 \u6027\u80fd\u66dd\u5149\uff0c\u5716\u50cf\u751f\u6210\u5feb4\u500d\uff0c\u6587\u672c\u63a8\u7406\u4ea6\u63d0\u5347\u8fd130%"},"content":{"rendered":"

\"news<\/p>\n

\n
\n
\n

Apple 在其 Machine Learning 研究博客上透露了 M5 處理器的最新性能表現。數據顯示，M5 不僅在文本推理方面較前代 M4 有顯著提升，更在圖形化生成任務中實現了數倍增長。<\/p>\n

<\/a>作為本次測試的核心平台，Apple 自研的 Machine Learning 框架 MLX 繼續發揮關鍵作用。該框架專為 Apple Silicon 設計，借助 RAM 架構，高效調度 CPU 與 GPU 資源，為大模型在 Mac 裝置上的流暢運行提供了基礎支撐。<\/p>\n

<\/a>在文本生成基準測試中，Apple 借助 MLX LM 工具包對多個主流開源大模型進行了評估，包括不同參數規格的 Qwen 系列與 GPT OSS 模型。結果顯示，M5 在生成後續文本 token 時的速度較 M4 提高了 19% 至 27%。<\/p>\n

<\/a>大語言模型的推理任務通常分為兩種類型：生成首個 token 的過程主要受限於計算能力，而持續生成後續內容則更依賴RAM 頻寬。因此 M5 憑借更高的 RAM 傳輸速率，在需要連續輸出文本的實際應用場景中表現更為出色。此外，搭載 24GB RAM 的 MacBook Pro 現已能夠流暢運行參數量達 300 億的 4-bit 量化混合專家模型（MoE），並將推理過程的內存佔用有效控制在 18GB 以內。<\/p>\n

除了文本處理，該處理器還集成了一款全新的 GPU 神經加速器，專門改善了 Machine Learning 任務中關鍵的矩陣乘法運算。在這一硬件升級的推動下，M5 執行圖像生成任務的速度達到了 M4 的 3.8 倍以上，為創意設計、視覺內容生成等應用提供了前所未有的流暢度和響應能力。<\/p>\n<\/div>\n<\/article>\n<\/div>\n

阅读更多（Read More）<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Apple 在其 Machine Learning 研&#31350 […]<\/p>\n","protected":false},"author":1,"featured_media":153466,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[24],"tags":[],"class_list":["post-153465","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-localnews"],"_links":{"self":[{"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/posts\/153465","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=153465"}],"version-history":[{"count":0,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/posts\/153465\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=\/wp\/v2\/media\/153466"}],"wp:attachment":[{"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=153465"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=153465"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.gocpmall.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=153465"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}