{"id":2684,"date":"2024-08-29T03:09:02","date_gmt":"2024-08-29T03:09:02","guid":{"rendered":"https:\/\/tastycounter.net\/index.php\/2024\/08\/29\/cerebras-ra-mat-cong-nghe-suy-luan-ai-nhanh-nhat-the-gioi-hieu-suat-gap-20-lan-so-voi-nvidia\/"},"modified":"2024-08-29T03:09:02","modified_gmt":"2024-08-29T03:09:02","slug":"cerebras-ra-mat-cong-nghe-suy-luan-ai-nhanh-nhat-the-gioi-hieu-suat-gap-20-lan-so-voi-nvidia","status":"publish","type":"post","link":"https:\/\/tastycounter.net\/index.php\/2024\/08\/29\/cerebras-ra-mat-cong-nghe-suy-luan-ai-nhanh-nhat-the-gioi-hieu-suat-gap-20-lan-so-voi-nvidia\/","title":{"rendered":"Cerebras ra m\u1eaft c\u00f4ng ngh\u1ec7 suy lu\u1eadn AI nhanh nh\u1ea5t th\u1ebf gi\u1edbi, hi\u1ec7u su\u1ea5t g\u1ea5p 20 l\u1ea7n so v\u1edbi NVIDIA"},"content":{"rendered":"<\/p>\n<div class=\"content-detail textview\">\n<div class=\"audio\"><audio controls><\/audio><\/div>\n<p>Cerebras Systems v\u1eeba ch\u00ednh th\u1ee9c c\u00f4ng b\u1ed1 Cerebras Inference, \u0111\u01b0\u1ee3c \u0111\u00e1nh gi\u00e1 l\u00e0 gi\u1ea3i ph\u00e1p suy lu\u1eadn AI nhanh nh\u1ea5t th\u1ebf gi\u1edbi. Cerebras Inference n\u00e0y cung c\u1ea5p hi\u1ec7u su\u1ea5t l\u00ean t\u1edbi 1.800 token m\u1ed7i gi\u00e2y cho m\u00f4 h\u00ecnh Llama 3.1 8B (8 t\u1ef7 tham s\u1ed1) v\u00e0 450 token m\u1ed7i gi\u00e2y cho Llama 3.1 70B, t\u1ee9c l\u00e0 nhanh h\u01a1n t\u1edbi g\u1ea7n 20 l\u1ea7n so v\u1edbi c\u00e1c gi\u1ea3i ph\u00e1p suy lu\u1eadn AI d\u1ef1a tr\u00ean GPU NVIDIA c\u00f3 s\u1eb5n trong c\u00e1c \u0111\u00e1m m\u00e2y quy m\u00f4 si\u00eau l\u1edbn hi\u1ec7n nay tr\u00ean to\u00e0n th\u1ebf gi\u1edbi, bao g\u1ed3m c\u1ea3 Microsoft Azure.<\/p>\n<p>Ngo\u00e0i hi\u1ec7u su\u1ea5t \u0111\u00e1ng kinh ng\u1ea1c, gi\u00e1 d\u1ecbch v\u1ee5 c\u1ee7a gi\u1ea3i ph\u00e1p suy lu\u1eadn m\u1edbi n\u00e0y c\u0169ng r\u1ea5t r\u1ebb, ch\u1ec9 b\u1eb1ng m\u1ed9t ph\u1ea7n nh\u1ecf so v\u1edbi c\u00e1c n\u1ec1n t\u1ea3ng \u0111\u00e1m m\u00e2y GPU ph\u1ed5 bi\u1ebfn. V\u00ed d\u1ee5: Kh\u00e1ch h\u00e0ng c\u00f3 th\u1ec3 nh\u1eadn \u0111\u01b0\u1ee3c m\u1ed9t tri\u1ec7u token ch\u1ec9 v\u1edbi 10 cent, do \u0111\u00f3 cung c\u1ea5p hi\u1ec7u su\u1ea5t gi\u00e1 cao h\u01a1n 100 l\u1ea7n cho kh\u1ed1i l\u01b0\u1ee3ng c\u00f4ng vi\u1ec7c AI.<\/p>\n<p>\u0110\u1ed9 ch\u00ednh x\u00e1c 16 bit v\u00e0 t\u1ed1c \u0111\u1ed9 suy lu\u1eadn nhanh h\u01a1n 20 l\u1ea7n c\u1ee7a Cerebras s\u1ebd cho ph\u00e9p nh\u00e0 ph\u00e1t tri\u1ec3n x\u00e2y d\u1ef1ng c\u00e1c \u1ee9ng d\u1ee5ng AI hi\u1ec7u su\u1ea5t cao th\u1ebf h\u1ec7 ti\u1ebfp theo m\u00e0 kh\u00f4ng ph\u1ea3i qu\u00e1 lo l\u1eb1ng v\u1ec1 t\u1ed1c \u0111\u1ed9 ho\u1eb7c chi ph\u00ed. T\u1ef7 l\u1ec7 gi\u00e1 th\u00e0nh\/hi\u1ec7u su\u1ea5t \u0111\u1ed9t ph\u00e1 n\u00e0y c\u00f3 th\u1ec3 th\u1ef1c hi\u1ec7n \u0111\u01b0\u1ee3c nh\u1edd h\u1ec7 th\u1ed1ng Cerebras CS-3 v\u00e0 b\u1ed9 x\u1eed l\u00fd AI Wafer Scale Engine 3 (WSE-3). CS-3 cung c\u1ea5p b\u0103ng th\u00f4ng b\u1ed9 nh\u1edb l\u1edbn h\u01a1n 7.000 l\u1ea7n so v\u1edbi Nvidia H100, gi\u1ea3i quy\u1ebft th\u00e1ch th\u1ee9c k\u1ef9 thu\u1eadt v\u1ec1 b\u0103ng th\u00f4ng b\u1ed9 nh\u1edb c\u1ee7a AI t\u1ea1o sinh.<\/p>\n<figure><img loading=\"lazy\" decoding=\"async\" src=\"https:\/\/st.quantrimang.com\/photos\/image\/holder.png\" width=\"760\" height=\"443\" class=\"lazy\" data-src=\"https:\/\/st.quantrimang.com\/photos\/image\/2024\/08\/28\/cerebras-ra-mat-cong-nghe-suy-luan-ai-nhanh-nhat-the-gioi1.jpg\"><\/figure>\n<div id=\"articleads\" class=\"adbox adsense in-article\"><ins class=\"adsbygoogle\" style=\"text-align:center\" data-ad-format=\"fluid\" data-ad-layout=\"in-article\" data-ad-client=\"ca-pub-9275417305531302\" data-ad-slot=\"2079243249\"><\/ins><\/div>\n<p>Cerebras Inference hi\u1ec7n kh\u1ea3 d\u1ee5ng \u1edf ba c\u1ea5p \u0111\u1ed9 sau:<\/p>\n<ul>\n<li>G\u00f3i mi\u1ec5n ph\u00ed Free Tier cung c\u1ea5p quy\u1ec1n truy c\u1eadp API mi\u1ec5n ph\u00ed v\u00e0 gi\u1edbi h\u1ea1n s\u1eed d\u1ee5ng h\u00e0o ph\u00f3ng cho b\u1ea5t k\u1ef3 ai \u0111\u0103ng nh\u1eadp.<\/li>\n<li>G\u00f3i Developer Tier \u0111\u01b0\u1ee3c thi\u1ebft k\u1ebf \u0111\u1ec3 tri\u1ec3n khai linh ho\u1ea1t, kh\u00f4ng c\u1ea7n m\u00e1y ch\u1ee7, cung c\u1ea5p cho ng\u01b0\u1eddi d\u00f9ng \u0111i\u1ec3m cu\u1ed1i API v\u1edbi chi ph\u00ed ch\u1ec9 b\u1eb1ng m\u1ed9t ph\u1ea7n nh\u1ecf so v\u1edbi c\u00e1c gi\u1ea3i ph\u00e1p thay th\u1ebf hi\u1ec7n c\u00f3 tr\u00ean th\u1ecb tr\u01b0\u1eddng, v\u1edbi c\u00e1c m\u1eabu Llama 3.1 8B v\u00e0 70B c\u00f3 gi\u00e1 l\u1ea7n l\u01b0\u1ee3t ch\u1ec9 l\u00e0 10 cent v\u00e0 60 cent cho m\u1ed9t tri\u1ec7u token.<\/li>\n<li>G\u00f3i Enterprise Tier cung c\u1ea5p c\u00e1c m\u00f4 h\u00ecnh tinh ch\u1ec9nh, th\u1ecfa thu\u1eadn m\u1ee9c d\u1ecbch v\u1ee5 t\u00f9y ch\u1ec9nh v\u00e0 h\u1ed7 tr\u1ee3 chuy\u00ean d\u1ee5ng. L\u00fd t\u01b0\u1edfng cho kh\u1ed1i l\u01b0\u1ee3ng c\u00f4ng vi\u1ec7c li\u00ean t\u1ee5c, doanh nghi\u1ec7p c\u00f3 th\u1ec3 truy c\u1eadp Cerebras Inference th\u00f4ng qua \u0111\u00e1m m\u00e2y ri\u00eang do Cerebras qu\u1ea3n l\u00fd ho\u1eb7c t\u1ea1i c\u01a1 s\u1edf c\u1ee7a kh\u00e1ch h\u00e0ng.<\/li>\n<\/ul>\n<blockquote>\n<p>V\u1edbi hi\u1ec7u su\u1ea5t k\u1ef7 l\u1ee5c, gi\u00e1 c\u1ea3 c\u1ea1nh tranh v\u00e0 quy\u1ec1n truy c\u1eadp API m\u1edf, Cerebras Inference \u0111\u1eb7t ra m\u1ed9t ti\u00eau chu\u1ea9n m\u1edbi cho vi\u1ec7c ph\u00e1t tri\u1ec3n v\u00e0 tri\u1ec3n khai LLM m\u1edf. L\u00e0 gi\u1ea3i ph\u00e1p duy nh\u1ea5t c\u00f3 kh\u1ea3 n\u0103ng cung c\u1ea5p c\u1ea3 \u0111\u00e0o t\u1ea1o v\u00e0 suy lu\u1eadn t\u1ed1c \u0111\u1ed9 cao, Cerebras m\u1edf ra nh\u1eefng kh\u1ea3 n\u0103ng ho\u00e0n to\u00e0n m\u1edbi cho AI.<\/p>\n<\/blockquote>\n<p>Trong b\u1ed1i c\u1ea3nh xu h\u01b0\u1edbng AI \u0111ang ph\u00e1t tri\u1ec3n nhanh ch\u00f3ng, v\u00e0 NVIDIA hi\u1ec7n \u0111ang n\u1eafm gi\u1eef v\u1ecb tr\u00ed th\u1ed1ng l\u0129nh tr\u00ean th\u1ecb tr\u01b0\u1eddng, s\u1ef1 xu\u1ea5t hi\u1ec7n c\u1ee7a c\u00e1c c\u00f4ng ty nh\u01b0 Cerebras v\u00e0 Groq b\u00e1o hi\u1ec7u m\u1ed9t s\u1ef1 thay \u0111\u1ed5i ti\u1ec1m n\u0103ng trong \u0111\u1ed9ng l\u1ef1c c\u1ee7a to\u00e0n ng\u00e0nh. Khi nhu c\u1ea7u v\u1ec1 c\u00e1c gi\u1ea3i ph\u00e1p suy lu\u1eadn AI nhanh h\u01a1n v\u00e0 ti\u1ebft ki\u1ec7m chi ph\u00ed h\u01a1n t\u0103ng l\u00ean, nh\u1eefng gi\u1ea3i ph\u00e1p nh\u01b0 Cerebras Inference \u0111ang \u1edf v\u1ecb th\u1ebf t\u1ed1t \u0111\u1ec3 n\u1eafm l\u1ea5y c\u01a1 h\u1ed9i ph\u00e1 v\u1ee1 s\u1ef1 th\u1ed1ng tr\u1ecb c\u1ee7a NVIDIA, \u0111\u1eb7c bi\u1ec7t l\u00e0 trong l\u0129nh v\u1ef1c suy lu\u1eadn.<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"<p>Cerebras Systems v\u1eeba ch\u00ednh th\u1ee9c c\u00f4ng b\u1ed1 Cerebras Inference, \u0111\u01b0\u1ee3c \u0111\u00e1nh gi\u00e1 l\u00e0 gi\u1ea3i ph\u00e1p suy lu\u1eadn AI nhanh nh\u1ea5t th\u1ebf gi\u1edbi. Cerebras Inference n\u00e0y cung c\u1ea5p hi\u1ec7u su\u1ea5t l\u00ean t\u1edbi 1.800 token m\u1ed7i gi\u00e2y cho m\u00f4 h\u00ecnh Llama 3.1 8B (8 t\u1ef7 tham s\u1ed1) v\u00e0 450 token m\u1ed7i gi\u00e2y cho Llama 3.1 70B, t\u1ee9c [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-2684","post","type-post","status-publish","format-standard","hentry","category-khong-phan-loai"],"_links":{"self":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/posts\/2684","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/comments?post=2684"}],"version-history":[{"count":0,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/posts\/2684\/revisions"}],"wp:attachment":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/media?parent=2684"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/categories?post=2684"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/tags?post=2684"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}