{"id":11475,"date":"2026-08-26T17:48:09","date_gmt":"2026-08-26T09:48:09","guid":{"rendered":"https:\/\/infernews.com\/blog\/openai-jalape-o-asic\/"},"modified":"2026-08-26T17:52:38","modified_gmt":"2026-08-26T09:52:38","slug":"openai-jalape-o-asic","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/openai-jalape-o-asic\/","title":{"rendered":"OpenAI Jalape\u00f1o\uff1a\u70ba\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u63a8\u7406\u800c\u9020\u7684 ASIC"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/https3A2F2Fsubstack-post-media.s3.amazonaws.com2Fpublic2Fimages2Fc028f756-4748-4583-91b2-5c4a63893cb5_2048x1153.jpg\" alt=\"Og image\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u7576\u6a21\u578b\u63a8\u7406\u6210\u672c\u53d7\u5230\u529f\u8017\u548c\u786c\u4ef6\u4f9b\u61c9\u9650\u5236\uff0c\u55ae\u9760\u901a\u7528 GPU \u672a\u5fc5\u80fd\u517c\u9867\u901f\u5ea6\u8207\u6548\u7387\u3002OpenAI \u4ee5\u7a7a\u767d\u8a2d\u8a08\u6253\u9020 Jalape\u00f1o\uff0c\u5c6c\u65bc\u5c08\u70ba\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff08Large Language Model\uff0cLLM\uff09\u63a8\u7406\u800c\u8a2d\u7684\u61c9\u7528\u5c08\u7528\u96c6\u6210\u96fb\u8def\uff08Application-Specific Integrated Circuit\uff0cASIC\uff09\uff0c\u4e26\u975e\u53ea\u91dd\u5c0d OpenAI \u81ea\u5bb6\u6a21\u578b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Jalape\u00f1o \u7531 OpenAI \u8207 Broadcom \u5408\u4f5c\u958b\u767c\uff0c\u8a2d\u8a08\u5de5\u4f5c\u65bc 2024 \u5e74\u4e2d\u958b\u59cb\uff0c\u7d04 <strong><mark>16 \u500b\u6708\u5f8c<\/mark><\/strong>\u5b8c\u6210\u88fd\u9020\u6d41\u7247\u3002\u6676\u7247\u63a1\u7528 HBM4 \u9ad8\u983b\u5bec\u8a18\u61b6\u9ad4\uff0c\u914d\u5408\u786c\u4ef6\u8207\u8edf\u4ef6\u5354\u540c\u8a2d\u8a08\uff0c\u76ee\u6a19\u662f\u5728\u4e0d\u540c\u6a21\u578b\u53ca\u63a8\u7406\u5de5\u4f5c\u8ca0\u8f09\u4e0b\u7dad\u6301\u9ad8\u6548\u80fd\uff0c\u800c\u975e\u904e\u5ea6\u512a\u5316\u55ae\u4e00\u63a8\u7406\u74b0\u7bc0\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5728 OpenAI \u5be6\u9a57\u5ba4\u4ee5 InferenceX \u6e2c\u8a66\u5957\u4ef6\u9032\u884c\u7684\u57fa\u6e96\u6e2c\u8a66\uff0c\u8072\u7a31 Jalape\u00f1o \u7684\u6bcf\u5146\u74e6\u541e\u5410\u91cf\u512a\u65bc\u5df2\u6e2c\u8a66\u7684 NVIDIA\u3001AMD \u53ca Google \u6676\u7247\uff0c\u4ea6\u5728\u6548\u80fd\u529f\u8017\u6bd4\u4e0a\u52dd\u904e Blackwell\u3002\u6bd4\u8f03\u6642 Jalape\u00f1o \u6c92\u6709\u4f7f\u7528 Multi Token Prediction\uff08MTP\uff09\uff0c\u5176\u4ed6\u6676\u7247\u5247\u63a1\u7528\u5404\u81ea\u6700\u4f73\u914d\u7f6e\uff0c\u56e0\u6b64\u7d50\u679c\u4ecd\u9700\u914d\u5408\u5b8c\u6574\u6e2c\u8a66\u689d\u4ef6\u7406\u89e3\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4ee5\u901a\u7528 AI \u63a8\u7406\u70ba\u8a2d\u8a08\u65b9\u5411\uff0c\u53ef\u904b\u884c\u591a\u985e\u6a21\u578b<\/li>\n\n\n\n<li>HBM4 \u4ee4\u8a18\u61b6\u9ad4\u983b\u5bec\u63a5\u8fd1\u9ad8\u968e GPU \u7684\u914d\u7f6e<\/li>\n\n\n\n<li>\u6838\u5fc3\u6307\u6a19\u805a\u7126\u6bcf\u5146\u74e6\u7684 token throughput<\/li>\n\n\n\n<li>\u6587\u7ae0\u672a\u63d0\u4f9b\u516c\u958b\u4e0b\u8f09\u3001\u5b89\u88dd\u6216\u4e00\u822c\u7528\u6236\u4f7f\u7528\u6d41\u7a0b<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u9019\u985e\u6676\u7247\u6700\u9069\u5408\u5927\u578b\u6a21\u578b\u670d\u52d9\u5546\u53ca\u8cc7\u6599\u4e2d\u5fc3\uff0c\u7528\u65bc\u63a8\u7406\u91cf\u5927\u3001\u96fb\u529b\u6210\u672c\u9ad8\u7684\u5de5\u4f5c\u6d41\u3002Jalape\u00f1o \u7684\u50f9\u503c\u5728\u65bc\u80fd\u5426\u628a\u9ad8\u541e\u5410\u91cf\u5ef6\u4f38\u81f3\u4e0d\u540c\u6a21\u578b\u548c\u9577\u671f\u904b\u884c\u74b0\u5883\uff0c\u800c\u4e0d\u53ea\u662f\u55ae\u6b21\u57fa\u6e96\u6e2c\u8a66\u4e2d\u7684\u512a\u52e2\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/newsletter.semianalysis.com\/p\/openai-jalapeno-better-than-nvidia\" target=\"_blank\" rel=\"noopener\">\u65b0\u805e\u4e3b\u9801<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>OpenAI \u5925\u62cd Broadcom \u6253\u9020 Jalape\u00f1o\uff0c\u7784\u6e96\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u63a8\u7406\uff0c\u4e26\u4ee5\u6bcf\u5146\u74e6\u541e\u5410\u91cf\u6311\u6230 NVIDIA GPU \u7684\u6548\u7387\u512a\u52e2\u3002<\/p>\n","protected":false},"author":8,"featured_media":11474,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,179,153,23],"tags":[],"class_list":["post-11475","post","type-post","status-publish","format-standard","hentry","category-133","category-nvidia","category-openai","category-news"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11475","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=11475"}],"version-history":[{"count":4,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11475\/revisions"}],"predecessor-version":[{"id":11481,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/11475\/revisions\/11481"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/11474"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=11475"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=11475"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=11475"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}