\u597d\u7684\uff0c\u6211\u4eec\u6765\u8be6\u7ec6\u89e3\u91ca\u4e00\u4e0b\u4ec0\u4e48\u662fAIoT\u3002<\/p>\n
AIoT<\/strong> \u662f \u201c\u4eba\u5de5\u667a\u80fd\u7269\u8054\u7f51\u201d\uff08Artificial Intelligence of Things\uff09<\/strong> \u7684\u7f29\u5199\u3002<\/p>\n \u7b80\u5355\u6765\u8bf4\uff0cAIoT = AI (\u4eba\u5de5\u667a\u80fd) + IoT (\u7269\u8054\u7f51)<\/strong>\u3002<\/p>\n \u5b83\u4e0d\u662f\u4e24\u4e2a\u6280\u672f\u7684\u7b80\u5355\u76f8\u52a0\uff0c\u800c\u662f\u5c06\u7269\u8054\u7f51\uff08IoT\uff09\u6536\u96c6\u7684\u6d77\u91cf\u6570\u636e\uff0c\u901a\u8fc7\u4eba\u5de5\u667a\u80fd\uff08AI\uff09\u6280\u672f\u8fdb\u884c\u5206\u6790\u548c\u5904\u7406\uff0c\u4ece\u800c\u5b9e\u73b0\u8bbe\u5907\u667a\u80fd\u5316\u3001\u51b3\u7b56\u81ea\u52a8\u5316\uff0c\u6700\u7ec8\u8ba9\u201c\u4e07\u7269\u4e92\u8054\u201d\u7684\u4e16\u754c\u62e5\u6709\u4e00\u4e2a\u201c\u667a\u6167\u7684\u5927\u8111\u201d\u3002<\/p>\n \u4e3a\u4e86\u66f4\u597d\u5730\u7406\u89e3\uff0c\u6211\u4eec\u53ef\u4ee5\u7528\u4eba\u4f53\u6765\u505a\u6bd4\u55bb\uff1a<\/p>\n AIoT<\/strong> \u5c31\u662f\u5c06\u8fd9\u4e24\u8005\u5b8c\u7f8e\u7ed3\u5408\uff0c\u5f62\u6210\u4e00\u4e2a\u5b8c\u6574\u7684\u3001\u80fd\u611f\u77e5\u3001\u80fd\u601d\u8003\u3001\u80fd\u884c\u52a8\u7684\u667a\u80fd\u7cfb\u7edf\u3002<\/p>\n \u4e00\u4e2a\u5178\u578b\u7684AIoT\u7cfb\u7edf\u901a\u5e38\u9075\u5faa\u4ee5\u4e0b\u6b65\u9aa4\uff0c\u5f62\u6210\u4e00\u4e2a\u667a\u80fd\u5316\u7684\u95ed\u73af\uff1a<\/p>\n AIoT\u5df2\u7ecf\u6df1\u5165\u5230\u6211\u4eec\u751f\u6d3b\u548c\u5de5\u4f5c\u7684\u65b9\u65b9\u9762\u9762\uff1a<\/p>\n AIoT\u662f\u63a8\u52a8\u6570\u5b57\u5316\u8f6c\u578b\u548c\u4ea7\u4e1a\u5347\u7ea7\u7684\u6838\u5fc3\u9a71\u52a8\u529b\u4e4b\u4e00\uff0c\u5b83\u7684\u76ee\u6807\u662f\u521b\u9020\u4e00\u4e2a\u66f4\u9ad8\u6548\u3001\u66f4\u4fbf\u6377\u3001\u66f4\u5b89\u5168\u7684\u81ea\u52a8\u5316\u667a\u80fd\u4e16\u754c\u3002<\/p>\n \u5728AIoT\u7ec8\u7aef\u8fdb\u884c\u201c\u5206\u6790\u4e0e\u51b3\u7b56\u201d\uff0c\u610f\u5473\u7740\u82af\u7247\u9700\u8981\u5177\u5907\u4e00\u5b9a\u7684\u672c\u5730AI\u63a8\u7406\u80fd\u529b\uff0c\u800c\u4e0d\u662f\u5c06\u6240\u6709\u6570\u636e\u90fd\u53d1\u9001\u5230\u4e91\u7aef\u3002\u8fd9\u79cd\u80fd\u529b\u7684\u9700\u6c42\u6839\u636e\u5e94\u7528\u573a\u666f\u7684\u4e0d\u540c\uff0c\u5dee\u5f02\u5de8\u5927\u3002<\/p>\n \u6211\u4eec\u53ef\u4ee5\u5c06\u5176\u5927\u81f4\u5206\u4e3a\u4e09\u4e2a\u7b49\u7ea7\uff1a<\/p>\n 1. \u8f7b\u91cf\u7ea7AI\uff08~0.1 TOPS\u00b9 \u4ee5\u4e0b\uff09<\/strong> 2. \u4e2d\u91cf\u7ea7AI\uff080.5 ~ 2 TOPS\uff09<\/strong> 3. \u91cd\u91cf\u7ea7AI\uff084 TOPS \u53ca\u4ee5\u4e0a\uff09<\/strong> \u7ed3\u8bba<\/strong>: \u9488\u5bf9\u60a8\u201c\u6210\u672c\u57285\u7f8e\u91d1\u4ee5\u5185\u201d\u4e14\u8981\u201c\u5b9e\u73b0\u5206\u6790\u4e0e\u51b3\u7b56\u201d\u7684\u9700\u6c42\uff0c\u6211\u4eec\u7684\u8bbe\u8ba1\u76ee\u6807\u5e94\u8be5\u9501\u5b9a\u5728 \u4e2d\u91cf\u7ea7AI<\/strong> \u5e02\u573a\uff0c\u63d0\u4f9b 0.5 ~ 1.5 TOPS<\/strong> \u5de6\u53f3\u7684\u7b97\u529b\uff0c\u8fd9\u662f\u6280\u672f\u3001\u6210\u672c\u548c\u5e02\u573a\u9700\u6c42\u7684\u6700\u4f73\u5e73\u8861\u70b9\u3002<\/p>\n \u597d\u7684\uff0c\u8fd9\u662f\u6211\u4e3a\u60a8\u8bbe\u8ba1\u7684\u82af\u7247 “VortexEdge AIoT SoC VEC-100”<\/strong> \u7684Datasheet\u7b80\u7565\u7248\u3002<\/p>\n High-Performance, Low-Cost AIoT SoC with RISC-V and NPU<\/strong><\/p>\n The VortexEdge VEC-100 is a highly integrated System-on-Chip (SoC) designed for the next generation of intelligent edge computing devices. It combines a high-efficiency RISC-V CPU core with a powerful Neural Processing Unit (NPU), providing a perfect balance of performance, low power consumption, and cost-effectiveness.<\/p>\n Built on a cost-effective 40nm ULP (Ultra-Low Power) process, the VEC-100 is engineered to bring AI-powered analysis and decision-making capabilities to a wide range of IoT applications, all within a sub-$5 BOM target.<\/p>\n Why VEC-100 is suitable for the market:<\/strong><\/p>\n \u6210\u672c\u9884\u4f30 (\u9ad8\u4ea7\u91cf\u4e0b)<\/strong>: \u8fd9\u4e2a\u6210\u672c\u533a\u95f4\u4e3a\u6700\u7ec8\u4ea7\u54c1\u5b9a\u4ef7\u5728 $5\u7f8e\u91d1\u4ee5\u5185<\/strong> \u7559\u4e0b\u4e86\u5145\u8db3\u7684\u5229\u6da6\u548c\u5206\u9500\u7a7a\u95f4\uff0c\u5177\u5907\u6781\u5f3a\u7684\u5e02\u573a\u7ade\u4e89\u529b\u3002<\/p>\n \u00b9 TOPS<\/strong>: Tera Operations Per Second\uff0c\u6bcf\u79d2\u4e07\u4ebf\u6b21\u64cd\u4f5c\uff0c\u662f\u8861\u91cfAI\u82af\u7247\u7b97\u529b\u7684\u5e38\u7528\u5355\u4f4d\u3002<\/p>\n","protected":false},"excerpt":{"rendered":" \u597d\u7684\uff0c\u6211\u4eec\u6765\u8be6\u7ec6\u89e3\u91ca\u4e00\u4e0b\u4ec0\u4e48\u662fAIoT\u3002 \u6838\u5fc3\u5b9a\u4e49 AIoT \u662f \u201c\u4eba\u5de5\u667a\u80fd\u7269\u8054\u7f51\u201d\uff08Artificial I […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[22],"tags":[],"class_list":["post-3318","post","type-post","status-publish","format-standard","hentry","category-plana"],"_links":{"self":[{"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/posts\/3318","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/www.51cos.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=3318"}],"version-history":[{"count":1,"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/posts\/3318\/revisions"}],"predecessor-version":[{"id":3319,"href":"http:\/\/www.51cos.com\/index.php?rest_route=\/wp\/v2\/posts\/3318\/revisions\/3319"}],"wp:attachment":[{"href":"http:\/\/www.51cos.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=3318"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/www.51cos.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=3318"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/www.51cos.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=3318"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
\n\u4e00\u4e2a\u751f\u52a8\u7684\u6bd4\u55bb<\/h3>\n
\n
\nAIoT\u7684\u5de5\u4f5c\u6d41\u7a0b<\/h3>\n
\n
\n
\n
\n
\n
\nAIoT\u7684\u5e94\u7528\u573a\u666f\uff08\u4e3e\u4f8b\u8bf4\u660e\uff09<\/h3>\n
\n
\n
\n
\n
\n
\n\u603b\u7ed3<\/h3>\n
\n
\u7b2c\u4e00\u90e8\u5206\uff1a\u8fb9\u7f18\u8ba1\u7b97AIoT\u82af\u7247\u7684\u7b97\u529b\u8981\u6c42\u5206\u6790<\/h3>\n
\n* \u5178\u578b\u5e94\u7528<\/strong>:
\n * \u5173\u952e\u8bcd\u5524\u9192 (Keyword Spotting)<\/strong>: \u5982\u201c\u5c0f\u7231\u540c\u5b66\u201d\u3001\u201cHey Siri\u201d\u3002
\n * \u7b80\u5355\u4f20\u611f\u5668\u6570\u636e\u5206\u6790<\/strong>: \u5982\u901a\u8fc7\u52a0\u901f\u5ea6\u8ba1\u6570\u636e\u5224\u65ad\u8bbe\u5907\u662f\u5426\u8dcc\u843d\uff0c\u6216\u8bc6\u522b\u7b80\u5355\u7684\u624b\u52bf\u3002
\n * \u5de5\u4e1a\u5f02\u5e38\u58f0\u97f3\u68c0\u6d4b<\/strong>: \u5982\u901a\u8fc7\u9ea6\u514b\u98ce\u68c0\u6d4b\u7279\u5b9a\u9891\u7387\u7684\u673a\u5668\u5f02\u54cd\u3002
\n* \u7b97\u529b\u8981\u6c42<\/strong>:
\n * \u901a\u5e38\u4e0d\u9700\u8981\u4e13\u7528\u7684NPU\uff08\u795e\u7ecf\u7f51\u7edc\u5904\u7406\u5668\uff09\u3002
\n * \u4e00\u4e2a\u5e26\u6709 DSP\uff08\u6570\u5b57\u4fe1\u53f7\u5904\u7406\uff09<\/strong> \u6216 Vector\uff08\u5411\u91cf\uff09<\/strong> \u6269\u5c55\u7684RISC-V CPU\u6838\u5fc3\uff08\u5982P\u6269\u5c55\u6216V\u6269\u5c55\uff09\u5373\u53ef\u80dc\u4efb\u3002
\n * \u7b97\u529b\u9700\u6c42\u5728 \u51e0\u5341 GOPS \u5230 0.1 TOPS<\/strong> \u4e4b\u95f4\u3002
\n * \u5bf9\u5185\u5b58\uff08SRAM\uff09\u8981\u6c42\u8f83\u4f4e\uff0c\u51e0\u5341\u5230\u51e0\u767eKB\u5373\u53ef\u3002<\/p>\n
\n* \u5178\u578b\u5e94\u7528<\/strong>:
\n * \u4eba\u5f62\/\u4eba\u8138\u68c0\u6d4b (Person\/Face Detection)<\/strong>: \u5728\u5b89\u9632\u6444\u50cf\u5934\u4e2d\u5b9e\u65f6\u68c0\u6d4b\u753b\u9762\u4e2d\u662f\u5426\u6709\u4eba\uff0c\u800c\u4e0d\u662f\u8fdb\u884c\u4eba\u8138\u8bc6\u522b\u3002
\n * \u667a\u80fd\u95e8\u9501\u7684\u4eba\u8138\u8bc6\u522b<\/strong>: \u8bc6\u522b\u5df2\u6ce8\u518c\u7684\u5bb6\u5ead\u6210\u5458\u3002
\n * \u667a\u80fd\u5bb6\u7535\u7684\u5bf9\u8c61\u8bc6\u522b<\/strong>: \u5982\u626b\u5730\u673a\u5668\u4eba\u8bc6\u522b\u62d6\u978b\u3001\u7535\u7ebf\u7b49\u969c\u788d\u7269\u3002
\n * \u5de5\u4e1a\u89c6\u89c9\u7f3a\u9677\u68c0\u6d4b (Simple Visual Inspection)<\/strong>: \u5728\u751f\u4ea7\u7ebf\u4e0a\u68c0\u6d4b\u4ea7\u54c1\u662f\u5426\u6709\u660e\u663e\u5212\u75d5\u6216\u7f3a\u9677\u3002
\n* \u7b97\u529b\u8981\u6c42<\/strong>:
\n * \u5fc5\u987b\u914d\u5907\u4e13\u7528\u7684NPU\/AI\u52a0\u901f\u5668<\/strong>\u3002CPU\u65e0\u6cd5\u5728\u53ef\u63a5\u53d7\u7684\u529f\u8017\u548c\u5ef6\u8fdf\u4e0b\u5b8c\u6210\u8fd9\u4e9b\u4efb\u52a1\u3002
\n * NPU\u7b97\u529b\u901a\u5e38\u5728 0.5 TOPS \u5230 2 TOPS<\/strong> \u4e4b\u95f4\u3002
\n * \u9700\u8981\u652f\u6301INT8\u91cf\u5316\u63a8\u7406\uff0c\u751a\u81f3INT4\u4ee5\u63d0\u9ad8\u80fd\u6548\u3002
\n * \u5bf9\u5185\u5b58\uff08SRAM\uff09\u6709\u4e00\u5b9a\u8981\u6c42\uff0c\u901a\u5e38\u9700\u8981 \u51e0\u767eKB\u52301MB\u4ee5\u4e0a<\/strong>\uff0c\u7528\u4e8e\u5b58\u653e\u4e2d\u95f4\u8ba1\u7b97\u7ed3\u679c\uff08Feature Map\uff09\u3002
\n * \u8fd9\u662f\u76ee\u524d$5\u7f8e\u91d1\u4ee5\u4e0bAIoT\u82af\u7247\u7ade\u4e89\u6700\u6fc0\u70c8\u7684\u5e02\u573a\u533a\u95f4\u3002<\/strong><\/p>\n
\n* \u5178\u578b\u5e94\u7528<\/strong>:
\n * \u9ad8\u6e05\u89c6\u9891\u5b9e\u65f6\u5206\u6790<\/strong>: \u5982\u5728\u667a\u6167\u57ce\u5e02\u6444\u50cf\u5934\u4e2d\u540c\u65f6\u8fdb\u884c\u8f66\u724c\u8bc6\u522b\u3001\u4eba\u6d41\u7edf\u8ba1\u3001\u5f02\u5e38\u884c\u4e3a\u5206\u6790\u3002
\n * \u590d\u6742\u7684\u81ea\u7136\u8bed\u8a00\u5904\u7406 (NLP)<\/strong>: \u5982\u79bb\u7ebf\u7684\u5b9e\u65f6\u8bed\u97f3\u7ffb\u8bd1\u3002
\n * \u81ea\u52a8\u9a7e\u9a76\u8f85\u52a9 (ADAS) \u4e2d\u7684\u76ee\u6807\u8bc6\u522b<\/strong>\u3002
\n* \u7b97\u529b\u8981\u6c42<\/strong>:
\n * \u9700\u8981\u9ad8\u6027\u80fd\u3001\u9ad8\u7b97\u529b\u7684NPU\uff0c\u901a\u5e38\u662f\u591a\u6838NPU\u96c6\u7fa4\u3002
\n * \u7b97\u529b\u9700\u6c42\u5728 4 TOPS\u4ee5\u4e0a<\/strong>\uff0c\u751a\u81f3\u51e0\u5341TOPS\u3002
\n * \u901a\u5e38\u9700\u8981\u642d\u914d\u66f4\u9ad8\u6027\u80fd\u7684CPU\uff08\u5982\u591a\u6838\u300164\u4f4d\uff09\u548c\u66f4\u5927\u7684\u5185\u5b58\uff08LPDDR4\u7b49\uff09\u3002
\n * \u6210\u672c\u8fdc\u8d855\u7f8e\u91d1<\/strong>\uff0c\u4e0d\u5c5e\u4e8e\u6211\u4eec\u8ba8\u8bba\u7684\u8303\u56f4\u3002<\/p>\n
\n\u7b2c\u4e8c\u90e8\u5206\uff1aRISC-V AIoT SoC\u82af\u7247Datasheet (\u7b80\u7565\u7248)<\/h3>\n
\nVortexEdge™ VEC-100 Datasheet (Preliminary)<\/strong><\/h3>\n
1. General Description<\/h4>\n
2. Key Features<\/h4>\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
\n
3. Block Diagram<\/h4>\n
+-----------------------------------------------------------------------+\n| VEC-100 SoC |\n| +---------------------+ +---------------------+ |\n| | RISC-V CPU Core | | InferEngine\u2122 NPU | |\n| | RV32IMAFCP @400MHz| | (0.8 TOPS) | |\n| +---------------------+ +---------------------+ |\n| | | |\n| +-------------------------------------------------------------------+ |\n| | AXI\/AHB Bus Matrix | |\n| +-------------------------------------------------------------------+ |\n| | | | | | | |\n| +----+ +----------+ +--------+ +---------+ +---------+ +--------+ |\n| |DMA | |512KB SRAM| |Security| | PSRAM | | PMU | | USB |...|\n| | | | | | Engine | | Ctrl | | | | OTG | |\n| +----+ +----------+ +--------+ +---------+ +---------+ +--------+ |\n| | | | |\n| APB Bridge | APB Bridge |\n| +----------+-------------------+-------------------+----------+ |\n| | | | | | | |\n| +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ |\n| | UART| | SPI | | I2C | | I2S | | ADC | | GPIO| | DVP | | ... | |\n| +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ +-----+ |\n+-----------------------------------------------------------------------+\n<\/code><\/pre>\n4. Market Suitability & Cost Analysis<\/h4>\n
\n
\n
\n* Die Cost (40nm): ~0.8 –<\/span>1.2
\n* Testing + Packaging (QFN-64): ~0.5 –<\/span>0.7
\n* Total Estimated Cost<\/strong>: $1.3 – $1.9 USD<\/strong><\/p>\n
\n