{"id":2188,"date":"2024-06-28T03:09:20","date_gmt":"2024-06-28T03:09:20","guid":{"rendered":"https:\/\/tastycounter.net\/index.php\/2024\/06\/28\/chip-ai-co-suc-manh-gap-20-lan-so-voi-nvidia-h100\/"},"modified":"2024-06-28T03:09:20","modified_gmt":"2024-06-28T03:09:20","slug":"chip-ai-co-suc-manh-gap-20-lan-so-voi-nvidia-h100","status":"publish","type":"post","link":"https:\/\/tastycounter.net\/index.php\/2024\/06\/28\/chip-ai-co-suc-manh-gap-20-lan-so-voi-nvidia-h100\/","title":{"rendered":"Chip AI c\u00f3 s\u1ee9c m\u1ea1nh g\u1ea5p 20 l\u1ea7n so v\u1edbi Nvidia H100"},"content":{"rendered":"<\/p>\n
\n

M\u1eabu chip AI Sohu m\u1edbi do Startup Etched, tr\u1ee5 s\u1edf t\u1ea1i San Francisco ph\u00e1t tri\u1ec3n c\u00f3 s\u1ee9c m\u1ea1nh g\u1ea5p 20 l\u1ea7n so v\u1edbi H100 c\u1ee7a Nvidia<\/a>.<\/p>\n

Theo Etched, Sohu l\u00e0 m\u1ed9t d\u1ea1ng m\u1ea1ch t\u00edch h\u1ee3p d\u00e0nh ri\u00eang cho \u1ee9ng d\u1ee5ng (ASIC) v\u00e0 c\u00f3 th\u1ec3 \u0111\u00e1nh b\u1ea1i GPU H100 c\u1ee7a Nvidia v\u1ec1 kh\u1ea3 n\u0103ng suy lu\u1eadn trong m\u00f4 h\u00ecnh ng\u00f4n ng\u1eef l\u1edbn (LLM). Hi\u1ec7u su\u1ea5t c\u1ee7a m\u1ed9t h\u1ec7 th\u1ed1ng v\u1edbi 8 chip Sohu c\u00f3 th\u1ec3 t\u01b0\u01a1ng \u0111\u01b0\u01a1ng 160 H100.<\/p>\n

\"Chip<\/p>\n

Etched cho bi\u1ebft, c\u00e1c h\u1ec7 th\u1ed1ng t\u0103ng t\u1ed1c AI hi\u1ec7n t\u1ea1i, d\u00f9 l\u00e0 CPU hay GPU, \u0111\u1ec1u \u0111\u01b0\u1ee3c thi\u1ebft k\u1ebf \u0111\u1ec3 ho\u1ea1t \u0111\u1ed9ng v\u1edbi c\u00e1c ki\u1ebfn tr\u00fac AI kh\u00e1c nhau nh\u01b0 M\u1ea1ng b\u1ed9 nh\u1edb ng\u1eafn h\u1ea1n (Short-term Memory Network), M\u1ea1ng th\u1ea7n kinh t\u00edch ch\u1eadp (Convolution Neural Network) hay m\u00f4 h\u00ecnh kh\u00f4ng gian tr\u1ea1ng th\u00e1i (State Space Model).<\/p>\n

Etched \u01b0\u1edbc t\u00ednh, ch\u1ec9 3,3% s\u1ed1 b\u00f3ng b\u00e1n d\u1eabn c\u1ee7a GPU H100 c\u1ee7a Nvidia \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng cho t\u00e1c v\u1ee5 ch\u00ednh, c\u00f2n 96,7% c\u00f2n l\u1ea1i \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng cho c\u00e1c t\u00e1c v\u1ee5 kh\u00e1c. V\u1edbi c\u00e1c chip AI \u0111a n\u0103ng \u0111i\u1ec1u n\u00e0y l\u00e0 c\u1ea7n thi\u1ebft nh\u01b0ng l\u1ea1i g\u00e2y ra s\u1ef1 l\u00e3ng ph\u00ed l\u1edbn.<\/p>\n

V\u00ec v\u1eady, c\u00f4ng ty kh\u1edfi nghi\u1ec7p Etched \u0111\u00e3 \u0111\u01b0a v\u00e0o Ki\u1ebfn tr\u00fac AI bi\u1ebfn \u00e1p (ki\u1ebfn tr\u00fac Transformer) v\u00e0o Sohu \u0111\u1ec3 c\u00f3 th\u1ec3 ph\u00e2n b\u1ed5 nhi\u1ec1u b\u00f3ng b\u00e1n d\u1eabn h\u01a1n cho t\u00ednh to\u00e1n tr\u00ed tu\u1ec7 nh\u00e2n t\u1ea1o tr\u00ean chip. H\u1ecd c\u1ed1 g\u1eafng t\u1ea1o ra chip \u0111\u01b0\u1ee3c t\u1ed1i \u01b0u h\u00f3a \u0111\u1ec3 ch\u1ea1y m\u1ed9t d\u1ea1ng m\u00f4 h\u00ecnh AI c\u1ee5 th\u1ec3 thay v\u00ec \u0111\u00e1p \u1ee9ng m\u1ecdi ki\u1ebfn tr\u00fac AI \u0111\u01a1n l\u1ebb. Th\u1ef1c t\u1ebf, ki\u1ebfn tr\u00fac Transformer c\u0169ng l\u00e0 n\u1ec1n t\u1ea3ng c\u1ee7a GPT (Generative Pre-trained Transformer), t\u01b0\u01a1ng t\u1ef1 v\u1edbi c\u00e1c h\u1ec7 th\u1ed1ng AI c\u1ee7a c\u00e1c h\u00e3ng c\u00f4ng ngh\u1ec7 l\u1edbn g\u1ed3m Sora, Gemini, Stable Diffusion v\u00e0 Dall-E.<\/p>\n

Etched cho bi\u1ebft, h\u1ecd \u0111\u00e3 \u0111\u1eb7t c\u01b0\u1ee3c v\u00e0o ki\u1ebfn tr\u00fac Transformer. N\u1ebfu Transformer bi\u1ebfn m\u1ea5t, c\u00f4ng ty kh\u1edfi nghi\u1ec7p n\u00e0y s\u1ebd ch\u1ebft nh\u01b0ng n\u1ebfu ti\u1ebfp t\u1ee5c t\u1ed3n t\u1ea1i Etched c\u00f3 ti\u1ec1m n\u0103ng tr\u1edf th\u00e0nh m\u1ed9t trong nh\u1eefng c\u00f4ng ty l\u1edbn nh\u1ea5t m\u1ecdi th\u1eddi.<\/p>\n

Gi\u00e1 c\u1ee7a Sohu ch\u01b0a \u0111\u01b0\u1ee3c c\u00f4ng b\u1ed1, c\u00f2n Nvidia H100 hi\u1ec7n c\u00f3 gi\u00e1 40.000 USD. Theo Tom’s Hardware, n\u1ebfu Sohu \u0111\u00e1p \u1ee9ng \u0111\u01b0\u1ee3c k\u1ef3 v\u1ecdng s\u1ebd gi\u00fap c\u00e1c trung t\u00e2m x\u1eed l\u00fd d\u1eef li\u1ec7u c\u00f3 th\u1ec3 ti\u1ebft ki\u1ec7m h\u00e0ng t\u1ef7 USD chi ph\u00ed ban \u0111\u1ea7u v\u00e0 chi ph\u00ed v\u1eadn h\u00e0nh. S\u1ea3n ph\u1ea9m n\u00e0y th\u1eadm ch\u00ed \u0111e d\u1ecda v\u1ecb th\u1ebf d\u1eabn \u0111\u1ea7u c\u1ee7a Nvidia nh\u01b0ng ch\u1ec9 khi chip c\u1ee7a Etched \u0111i v\u00e0o v\u1eadn h\u00e0nh.<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

M\u1eabu chip AI Sohu m\u1edbi do Startup Etched, tr\u1ee5 s\u1edf t\u1ea1i San Francisco ph\u00e1t tri\u1ec3n c\u00f3 s\u1ee9c m\u1ea1nh g\u1ea5p 20 l\u1ea7n so v\u1edbi H100 c\u1ee7a Nvidia. Theo Etched, Sohu l\u00e0 m\u1ed9t d\u1ea1ng m\u1ea1ch t\u00edch h\u1ee3p d\u00e0nh ri\u00eang cho \u1ee9ng d\u1ee5ng (ASIC) v\u00e0 c\u00f3 th\u1ec3 \u0111\u00e1nh b\u1ea1i GPU H100 c\u1ee7a Nvidia v\u1ec1 kh\u1ea3 n\u0103ng suy lu\u1eadn […]<\/p>\n","protected":false},"author":1,"featured_media":2189,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-2188","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-khong-phan-loai"],"_links":{"self":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/posts\/2188","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=2188"}],"version-history":[{"count":0,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/posts\/2188\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/media\/2189"}],"wp:attachment":[{"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/media?parent=2188"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/categories?post=2188"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/tastycounter.net\/index.php\/wp-json\/wp\/v2\/tags?post=2188"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}