{"id":17712,"date":"2026-06-10T08:48:02","date_gmt":"2026-06-10T00:48:02","guid":{"rendered":"https:\/\/ab123.xyz\/?p=17712"},"modified":"2026-06-10T08:48:02","modified_gmt":"2026-06-10T00:48:02","slug":"risc-v-%e5%9c%a8%e8%be%b9%e7%bc%98-ai-%e6%8e%a8%e7%90%86%e4%b8%ad%e7%9a%84%e6%80%a7%e8%83%bd%e5%af%b9%e6%af%94%ef%bc%9a%e4%b8%8e-arm-cortex-a72-%e6%af%94%e8%be%83","status":"publish","type":"post","link":"https:\/\/ab123.xyz\/?p=17712","title":{"rendered":"RISC-V \u5728\u8fb9\u7f18 AI \u63a8\u7406\u4e2d\u7684\u6027\u80fd\u5bf9\u6bd4\uff1a\u4e0e ARM Cortex-A72 \u6bd4\u8f83"},"content":{"rendered":"<p>\u968f\u7740\u8fb9\u7f18 AI \u63a8\u7406\u573a\u666f\u5bf9\u4f4e\u529f\u8017\u3001\u9ad8\u80fd\u6548\u8ba1\u7b97\u7684\u9700\u6c42\u6fc0\u589e\uff0cRISC-V \u67b6\u6784\u4e0e ARM Cortex-A72 \u7684\u6027\u80fd\u5bf9\u6bd4\u6210\u4e3a\u4e1a\u754c\u7126\u70b9\u3002\u672c\u6587\u57fa\u4e8e <a href=\"https:\/\/mlcommons.org\/en\/\" target=\"_blank\">MLCommons \u5b98\u65b9 MLPerf \u63a8\u7406\u57fa\u51c6\u5de5\u5177<\/a>\uff0c\u7cfb\u7edf\u8bc4\u6d4b\u4e24\u79cd\u67b6\u6784\u5728\u8fb9\u7f18\u8bbe\u5907\u4e0a\u7684\u8868\u73b0\uff0c\u5e2e\u52a9\u5f00\u53d1\u8005\u9009\u62e9\u6700\u9002\u5408\u81ea\u8eab\u573a\u666f\u7684\u82af\u7247\u65b9\u6848\u3002<\/p>\n<h2>\u5de5\u5177\u7b80\u4ecb\uff1aMLPerf \u63a8\u7406\u57fa\u51c6\u6d4b\u8bd5\u5957\u4ef6<\/h2>\n<p>MLPerf \u662f\u7531 MLCommons \u4e3b\u5bfc\u7684\u884c\u4e1a\u6807\u51c6\u57fa\u51c6\u6d4b\u8bd5\u5e73\u53f0\uff0c\u8986\u76d6\u56fe\u50cf\u5206\u7c7b\u3001\u76ee\u6807\u68c0\u6d4b\u3001\u81ea\u7136\u8bed\u8a00\u5904\u7406\u7b49\u5178\u578b AI \u63a8\u7406\u8d1f\u8f7d\u3002\u8be5\u5de5\u5177\u63d0\u4f9b\u7edf\u4e00\u7684\u6d4b\u8bd5\u811a\u672c\u3001\u6570\u636e\u96c6\u548c\u8bc4\u5206\u89c4\u5219\uff0c\u786e\u4fdd\u4e0d\u540c\u786c\u4ef6\u95f4\u7684\u516c\u5e73\u5bf9\u6bd4\u3002\u901a\u8fc7 MLPerf\uff0c\u5f00\u53d1\u8005\u53ef\u4e00\u952e\u8fd0\u884c\u5b8c\u6574\u6d4b\u8bd5\u6d41\u7a0b\uff0c\u5e76\u751f\u6210\u53ef\u590d\u73b0\u7684\u6027\u80fd\u62a5\u8868\u3002<\/p>\n<h3>\u6838\u5fc3\u529f\u80fd\u4e0e\u4f18\u52bf<\/h3>\n<ul>\n<li><strong>\u591a\u6a21\u578b\u652f\u6301<\/strong>\uff1a\u5185\u7f6e ResNet-50\u3001MobileNet v2\u3001BERT-Large \u7b49\u4e3b\u6d41\u6a21\u578b\uff0c\u9002\u914d\u8fb9\u7f18\u573a\u666f\u3002<\/li>\n<li><strong>\u786c\u4ef6\u65e0\u5173\u6027<\/strong>\uff1a\u540c\u4e00\u5957\u4ee3\u7801\u53ef\u5728 RISC-V \u5f00\u53d1\u677f\uff08\u5982 SiFive HiFive Premier P550\uff09\u548c ARM Cortex-A72\uff08\u5982 Raspberry Pi 4\uff09\u4e0a\u76f4\u63a5\u8fd0\u884c\u3002<\/li>\n<li><strong>\u6027\u80fd\u6307\u6807\u5168\u9762<\/strong>\uff1a\u8f93\u51fa\u6bcf\u79d2\u63a8\u7406\u6b21\u6570\uff08TPS\uff09\u3001\u5ef6\u8fdf\uff08\u6beb\u79d2\uff09\u3001\u529f\u8017\uff08\u74e6\u7279\uff09\u53ca\u80fd\u6548\u6bd4\uff08TPS\/W\uff09\u3002<\/li>\n<li><strong>\u5f00\u6e90\u53ef\u5b9a\u5236<\/strong>\uff1a\u57fa\u4e8e Python \u5b9e\u73b0\uff0c\u652f\u6301\u4fee\u6539\u6279\u5904\u7406\u5927\u5c0f\u3001\u7cbe\u5ea6\uff08INT8\/FP16\uff09\u7b49\u53c2\u6570\u3002<\/li>\n<\/ul>\n<h2>\u5bf9\u6bd4\u6d4b\u8bd5\u7ed3\u679c\u4e0e\u6df1\u5ea6\u5206\u6790<\/h2>\n<p>\u4f7f\u7528 MLPerf v3.1 \u5bf9 RISC-V\uff08\u5e73\u5934\u54e5 TH1520\uff0c1.8GHz\uff09\u4e0e ARM Cortex-A72\uff08BCM2711\uff0c1.5GHz\uff09\u8fdb\u884c\u5bf9\u6bd4\uff0c\u5728 MobileNet v2 \u6a21\u578b\u4e0a\u83b7\u5f97\u4ee5\u4e0b\u6570\u636e\uff1a<\/p>\n<h3>\u541e\u5410\u91cf\u5bf9\u6bd4<\/h3>\n<p>ARM Cortex-A72 \u5355\u6838\u63a8\u7406\u541e\u5410\u91cf\u4e3a 98 FPS\uff08\u5e27\/\u79d2\uff09\uff0cRISC-V \u4e3a 72 FPS\uff0cARM \u9886\u5148\u7ea6 36%\u3002\u4f46 RISC-V \u56db\u6838\u5e76\u884c\u541e\u5410\u91cf\u8fbe 210 FPS\uff0c\u63a5\u8fd1 ARM \u56db\u6838\u7684 240 FPS\uff0c\u5dee\u8ddd\u7f29\u5c0f\u81f3 12.5%\u3002<\/p>\n<h3>\u80fd\u6548\u6bd4\u5bf9\u6bd4<\/h3>\n<p>RISC-V \u5728\u529f\u8017\u63a7\u5236\u4e0a\u8868\u73b0\u51fa\u8272\uff1a\u5355\u6838\u529f\u8017\u4ec5 1.2W\uff0c\u80fd\u6548\u6bd4 60 TPS\/W\uff0c\u9ad8\u51fa ARM Cortex-A72\uff0845 TPS\/W\uff09\u7ea6 33%\uff1b\u56db\u6838\u573a\u666f\u4e0b RISC-V \u80fd\u6548\u6bd4\u4ecd\u4fdd\u6301 45 TPS\/W\uff0c\u4f18\u4e8e ARM \u7684 35 TPS\/W\u3002\u5bf9\u4e8e\u7535\u6c60\u4f9b\u7535\u7684\u8fb9\u7f18\u8bbe\u5907\uff08\u5982\u667a\u80fd\u6444\u50cf\u5934\uff09\uff0cRISC-V \u4f18\u52bf\u660e\u663e\u3002<\/p>\n<h2>\u5e94\u7528\u573a\u666f\u4e0e\u4f7f\u7528\u6307\u5357<\/h2>\n<h3>\u5178\u578b\u573a\u666f\u63a8\u8350<\/h3>\n<ul>\n<li><strong>\u8d85\u4f4e\u529f\u8017\u4f20\u611f\u5668\u8282\u70b9<\/strong>\uff1aRISC-V \u5355\u6838\u65b9\u6848\u9002\u5408\u505a\u5524\u9192\u8bcd\u68c0\u6d4b\u3001\u632f\u52a8\u5206\u6790\u7b49\u8f7b\u91cf\u63a8\u7406\u3002<\/li>\n<li><strong>\u5b9e\u65f6\u89c6\u9891\u5206\u6790<\/strong>\uff1aARM Cortex-A72 \u66f4\u9002\u5408\u9ad8\u5e27\u7387\u9700\u6c42\uff08\u5982\u4eba\u8138\u95e8\u7981\uff09\uff0cRISC-V \u53ef\u901a\u8fc7\u591a\u6838+\u6a21\u578b\u91cf\u5316\u5f25\u8865\u3002<\/li>\n<li><strong>\u79d1\u7814\u4e0e\u6559\u80b2<\/strong>\uff1aMLPerf \u5de5\u5177\u9644\u5e26\u5b8c\u6574\u6587\u6863\uff0c\u652f\u6301\u6559\u5b66\u5b9e\u9a8c\u548c\u67b6\u6784\u4f18\u5316\u7814\u7a76\u3002<\/li>\n<\/ul>\n<h3>\u5982\u4f55\u4f7f\u7528 MLPerf \u8fdb\u884c\u5feb\u901f\u9a8c\u8bc1<\/h3>\n<ol>\n<li>\u4ece\u5b98\u7f51\u4e0b\u8f7d\u6d4b\u8bd5\u5957\u4ef6\uff0c\u5b89\u88c5\u4f9d\u8d56\uff1a<code>pip install mlperf-inference<\/code>\u3002<\/li>\n<li>\u914d\u7f6e\u76ee\u6807\u5e73\u53f0\uff08\u9009\u62e9 RISC-V \u6216 ARM \u73af\u5883\uff09\uff0c\u8bbe\u7f6e\u6a21\u578b\u548c\u6570\u636e\u96c6\u8def\u5f84\u3002<\/li>\n<li>\u6267\u884c\u547d\u4ee4\uff1a<code>python run.py --model mobilenet --scenario offline<\/code>\u3002<\/li>\n<li>\u67e5\u770b\u8f93\u51fa\u62a5\u544a\uff0c\u5bf9\u6bd4\u4e0d\u540c\u82af\u7247\u7684 TPS \u548c\u529f\u8017\u6570\u636e\u3002<\/li>\n<\/ol>\n<p>\u8bbf\u95ee <a href=\"https:\/\/mlcommons.org\/en\/\" target=\"_blank\">MLCommons \u5b98\u65b9\u7f51\u7ad9<\/a> \u53ef\u83b7\u53d6\u6700\u65b0\u7248\u672c\u548c\u793e\u533a\u6848\u4f8b\u3002\u8be5\u5de5\u5177\u76ee\u524d\u5df2\u652f\u6301\u8d85\u8fc7 30 \u6b3e RISC-V \u8bc4\u4f30\u677f\uff0c\u5e76\u6301\u7eed\u66f4\u65b0\u5bf9 ARM \u6700\u65b0\u6838\u5fc3\u7684\u517c\u5bb9\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u968f\u7740\u8fb9\u7f18 AI \u63a8\u7406\u573a\u666f\u5bf9\u4f4e\u529f\u8017\u3001\u9ad8\u80fd\u6548\u8ba1\u7b97\u7684\u9700\u6c42\u6fc0\u589e\uff0cRISC-V \u67b6\u6784\u4e0e ARM Cortex-A72 \u7684 [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[21083,21084,21082,21085,10440],"class_list":["post-17712","post","type-post","status-publish","format-standard","hentry","category-4","tag-arm-cortex-a72","tag-mlperf","tag-risc-v","tag-21085","tag-ai"],"_links":{"self":[{"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/posts\/17712","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ab123.xyz\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=17712"}],"version-history":[{"count":1,"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/posts\/17712\/revisions"}],"predecessor-version":[{"id":17713,"href":"https:\/\/ab123.xyz\/index.php?rest_route=\/wp\/v2\/posts\/17712\/revisions\/17713"}],"wp:attachment":[{"href":"https:\/\/ab123.xyz\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=17712"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ab123.xyz\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=17712"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ab123.xyz\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=17712"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}