{"id":49540,"date":"2023-11-14T20:08:00","date_gmt":"2023-11-14T12:08:00","guid":{"rendered":"https:\/\/www.ioiotimes.com\/?p=49540"},"modified":"2023-11-15T09:09:24","modified_gmt":"2023-11-15T01:09:24","slug":"nvidia%e5%a2%9e%e5%bc%b7%e4%b8%96%e7%95%8c%e9%a0%98%e5%85%88%e7%9a%84%e4%ba%ba%e5%b7%a5%e6%99%ba%e6%85%a7%e9%81%8b%e7%ae%97%e5%b9%b3%e5%8f%b0hopper%ef%bc%8c%e5%85%a8%e7%90%83%e9%a0%82%e7%b4%9a","status":"publish","type":"post","link":"https:\/\/www.ioiotimes.com\/?p=49540","title":{"rendered":"NVIDIA\u589e\u5f37\u4e16\u754c\u9818\u5148\u7684\u4eba\u5de5\u667a\u6167\u904b\u7b97\u5e73\u53f0Hopper\uff0c\u5168\u7403\u9802\u7d1a\u4f3a\u670d\u5668\u88fd\u9020\u5546\u548c\u96f2\u7aef\u670d\u52d9\u4f9b\u61c9\u5546\u5373\u5c07\u63a8\u51faHGX H200\u7cfb\u7d71\u548c\u96f2\u7aef\u57f7\u884c\u500b\u9ad4"},"content":{"rendered":"\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"576\" src=\"https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2023\/11\/20231114-nvidia02.jpg\" alt=\"\" class=\"wp-image-49547\" title=\"\" srcset=\"https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2023\/11\/20231114-nvidia02.jpg 1024w, https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2023\/11\/20231114-nvidia02-300x169.jpg 300w, https:\/\/www.ioiotimes.com\/wordpress\/wp-content\/uploads\/2023\/11\/20231114-nvidia02-768x432.jpg 768w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p><\/p>\n\n\n\n<p>NVIDIA (NASDAQ: NVDA) \u4eca\u5929\u5ba3\u5e03\u63a8\u51fa NVIDIA HGX<ins>\u2122<\/ins> H200\uff0c\u70ba\u5168\u7403\u9818\u5148\u7684\u4eba\u5de5\u667a\u6167\u904b\u7b97\u5e73\u53f0\u5e36\u4f86\u5f37\u5927\u52d5\u529b\u3002\u8a72\u5e73\u53f0\u57fa\u65bc NVIDIA Hopper\u2122 \u67b6\u69cb\uff0c\u914d\u5099 NVIDIA H200 Tensor Core GPU\uff0c\u4e26\u5177\u6709\u5148\u9032\u7684\u8a18\u61b6\u9ad4\uff0c\u53ef\u70ba\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u548c\u9ad8\u6548\u80fd\u904b\u7b97\u7684\u5de5\u4f5c\u8ca0\u8f09\u8655\u7406\u5927\u91cf\u8cc7\u6599\u3002<\/p>\n\n\n\n<p>NVIDIA H200 \u662f\u9996\u6b3e\u63d0\u4f9b HBM3e \u7684 GPU\uff0cHBM3e \u662f\u901f\u5ea6\u66f4\u5feb\u3001\u5bb9\u91cf\u66f4\u5927\u7684\u8a18\u61b6\u9ad4\uff0c\u53ef\u52a0\u901f\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u548c\u5927\u578b\u8a9e\u8a00\u6a21\u578b\uff0c\u540c\u6642\u63a8\u9032\u9ad8\u6548\u80fd\u904b\u7b97\u5de5\u4f5c\u8ca0\u8f09\u7684\u79d1\u5b78\u904b\u7b97\u3002\u501f\u52a9 HBM3e\uff0cNVIDIA H200 \u4ee5\u6bcf\u79d2 4.8 TB \u7684\u901f\u5ea6\u63d0\u4f9b 141GB \u8a18\u61b6\u9ad4\uff0c\u8207\u5176\u524d\u4ee3\u7522\u54c1 NVIDIA A100 \u76f8\u6bd4\uff0c\u5bb9\u91cf\u5e7e\u4e4e\u7ffb\u500d\uff0c\u983b\u5bec\u589e\u52a0 2.4 \u500d\u3002<\/p>\n\n\n\n<p>\u5168\u7403\u9818\u5148\u4f3a\u670d\u5668\u88fd\u9020\u5546\u548c\u96f2\u7aef\u670d\u52d9\u4f9b\u61c9\u5546\u63a1\u7528 H200 \u7684\u7cfb\u7d71\u9810\u8a08\u5c07\u65bc 2024 \u5e74\u7b2c\u4e8c\u5b63\u958b\u59cb\u51fa\u8ca8\u3002<\/p>\n\n\n\n<p>NVIDIA \u8d85\u5927\u898f\u6a21\u8207\u9ad8\u6548\u80fd\u904b\u7b97\u90e8\u9580\u526f\u7e3d\u88c1 Ian Buck \u8868\u793a\uff1a\u300c\u8981\u900f\u904e\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u548c\u9ad8\u6548\u80fd\u904b\u7b97\u61c9\u7528\u7a0b\u5f0f\u5275\u9020\u667a\u6167\uff0c\u5fc5\u9808\u4f7f\u7528\u5927\u578b\u3001\u5feb\u901f\u7684 GPU \u8a18\u61b6\u9ad4\u4f86\u9ad8\u901f\u6709\u6548\u5730\u8655\u7406\u5927\u91cf\u8cc7\u6599\u3002\u6191\u85c9 NVIDIA H200\uff0c\u696d\u754c\u9818\u5148\u7684\u7aef\u5230\u7aef\u4eba\u5de5\u667a\u6167\u8d85\u7d1a\u904b\u7b97\u5e73\u53f0\u901f\u5ea6\u8b8a\u5f97\u66f4\u5feb\uff0c\u80fd\u5920\u89e3\u6c7a\u4e16\u754c\u4e0a\u4e00\u4e9b\u6700\u91cd\u8981\u7684\u6311\u6230\u3002\u300d<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><strong>\u6301\u7e8c\u7684\u5275\u65b0\uff0c\u6301\u7e8c\u7684\u6548\u80fd\u8e8d\u9032<\/strong><\/p>\n\n\n\n<p>NVIDIA Hopper \u67b6\u69cb\u8207\u5176\u524d\u4ee3\u67b6\u69cb\u76f8\u6bd4\uff0c\u5be6\u73fe\u4e86\u524d\u6240\u672a\u6709\u7684\u6548\u80fd\u8e8d\u9032\uff0c\u4e26\u900f\u904e H100 \u7684\u6301\u7e8c\u8edf\u9ad4\u589e\u5f37\uff0c\u5305\u62ec\u50cf\u662f\u6700\u8fd1\u91cb\u51fa <a href=\"https:\/\/developer.nvidia.com\/tensorrt\" target=\"_blank\" rel=\"noreferrer noopener\">NVIDIA TensorRT\u2122-LLM<\/a> \u9019\u7a2e\u5f37\u5927\u7684\u958b\u6e90\u5eab\uff0c\u4e0d\u65b7\u63d0\u9ad8\u6548\u80fd\u6c34\u6e96\u3002<\/p>\n\n\n\n<p>H200 \u7684\u63a8\u51fa\u5c07\u5e36\u4f86\u66f4\u9032\u4e00\u6b65\u7684\u6548\u80fd\u8e8d\u9032\uff0c\u5305\u62ec\u76f8\u5c0d\u65bc H100\uff0c\u5728 Llama 2 \u9019\u6a23\u4e00\u500b\u64c1\u6709 700 \u5104\u53c3\u6578\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u4e0a\u7684\u63a8\u8ad6\u901f\u5ea6\u5c07\u8fd1\u7ffb\u500d\u63d0\u5347\u3002\u672a\u4f86\u7684\u8edf\u9ad4\u66f4\u65b0\u9810\u8a08\u5c07\u70ba H200 \u5e36\u4f86\u984d\u5916\u6548\u80fd\u7684\u9818\u5148\u512a\u52e2\uff0c\u4e26\u6301\u7e8c\u6700\u4f73\u5316 H200\u3002<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><strong>NVIDIA H200<\/strong><strong>\u7684\u5916\u578b\u8207\u898f\u683c<\/strong><\/p>\n\n\n\n<p>NVIDIA H200 \u5c07\u63d0\u4f9b\u5305\u542b\u5177\u6709\u56db\u8def\u548c\u516b\u8def\u914d\u7f6e\u7684 NVIDIA HGX H200 \u4f3a\u670d\u5668\u4e3b\u6a5f\u677f\uff0c\u5176\u8edf\u786c\u9ad4\u7686\u8207 HGX 100 \u7cfb\u7d71\u76f8\u5bb9\u3002\u6b64\u5916\uff0cNVIDIA H200 \u4ea6\u53ef\u8207\u4eca\u5e74 8 \u6708\u63a8\u51fa\u3001\u63a1\u7528 HBM3e \u7684 <a href=\"https:\/\/blogs.nvidia.com.tw\/2023\/08\/09\/gh200-grace-hopper-superchip-with-hbm3e-memory\/\" target=\"_blank\" rel=\"noreferrer noopener\">NVIDIA GH200 Grace Hopper\u2122<\/a> \u8d85\u7d1a\u6676\u7247\u642d\u914d\u4f7f\u7528\u3002<\/p>\n\n\n\n<p>\u900f\u904e\u9019\u4e9b\u9078\u9805\uff0cH200 \u53ef\u4ee5\u90e8\u7f72\u5728\u5404\u7a2e\u985e\u578b\u7684\u8cc7\u6599\u4e2d\u5fc3\u4e2d\uff0c\u5305\u62ec\u672c\u5730\u3001\u96f2\u7aef\u3001\u6df7\u5408\u96f2\u548c\u908a\u7de3\u74b0\u5883\u3002 NVIDIA \u5168\u7403\u751f\u614b\u7cfb\u5408\u4f5c\u5925\u4f34\u4f3a\u670d\u5668\u88fd\u9020\u5546\uff0c\u5305\u62ec<ins><a href=\"https:\/\/www.asrockrack.com\/general\/news.asp?id=226\" target=\"_blank\" rel=\"noreferrer noopener\">\u6c38\u64ce\u96fb\u5b50<\/a><\/ins>\u3001<ins><a href=\"https:\/\/server.asus.com\/news\/ASUS-Demonstrates-AI-and-Immersion-Cooling-Solutions-at-SC23\" target=\"_blank\" rel=\"noreferrer noopener\">\u83ef\u78a9<\/a><\/ins>\u3001\u6234\u723e\u79d1\u6280\u96c6\u5718\u3001Eviden\u3001<ins><a href=\"https:\/\/www.gigabyte.com\/Press\/News\/2124\" target=\"_blank\" rel=\"noreferrer noopener\">\u6280\u5609\u79d1\u6280<\/a><\/ins>\u3001\u6167\u8207\u79d1\u6280\u3001<ins><a href=\"https:\/\/www.ingrasys.com\/gb6181\" target=\"_blank\" rel=\"noreferrer noopener\">\u9d3b\u4f70\u79d1\u6280<\/a><\/ins>\u3001\u806f\u60f3\u3001<ins><a href=\"https:\/\/www.qct.io\/Press-Releases\/index\/PR\/Server\/QCT-Brings-GPU-Powerhouses-Powered-by-NVIDIA-to-SC23\" target=\"_blank\" rel=\"noreferrer noopener\">\u96f2\u9054\u79d1\u6280<\/a><\/ins>\u3001\u7f8e\u8d85\u5fae\u3001\u7def\u5275\u8cc7\u901a\u548c\u7def\u7a4e\u79d1\u6280\uff0c\u90fd\u53ef\u4ee5\u8f15\u9b06\u4f7f\u7528\u57fa\u65bc H200 \u7684\u76f4\u63a5\u66ff\u63db\u65b9\u5f0f\u4f86\u66f4\u65b0\u5176\u57fa\u65bc H100 \u7684\u7cfb\u7d71\u3002<\/p>\n\n\n\n<p>\u9664\u4e86 <ins><a href=\"https:\/\/www.coreweave.com\/blog\/kicking-off-sc23-coreweave-collaborates-with-nvidia\" target=\"_blank\" rel=\"noreferrer noopener\">CoreWeave<\/a><\/ins>\u3001<ins><a href=\"https:\/\/lambdalabs.com\/blog\/lambda-cloud-clusters-to-support-nvidia-h200-tensor-core-gpus?utm_medium=press-release&amp;utm_campaign=2023-11-NVIDIA-announcements\" target=\"_blank\" rel=\"noreferrer noopener\">Lambda<\/a><\/ins> \u548c Vultr \u4e4b\u5916\uff0c\u5f9e\u660e\u5e74\u958b\u59cb\uff0c\u4e9e\u99ac\u905c\u7db2\u8def\u670d\u52d9\u3001Google Cloud\u3001Microsoft Azure \u548c Oracle Cloud Infrastructure \u4e5f\u5c07\u6210\u70ba\u7b2c\u4e00\u6279\u90e8\u7f72\u57fa\u65bc H200 \u57f7\u884c\u500b\u9ad4\u7684\u96f2\u7aef\u670d\u52d9\u4f9b\u61c9\u5546\u3002<\/p>\n\n\n\n<p>HGX H200 \u7531 NVIDIA\u00ae NVLink\u2122 \u548c NVSwitch\u2122 \u9ad8\u901f\u4e92\u9023\u6280\u8853\u9a45\u52d5\uff0c\u53ef\u70ba\u5404\u7a2e\u61c9\u7528\u5de5\u4f5c\u8ca0\u8f09\u63d0\u4f9b\u6700\u9ad8\u6548\u80fd\uff0c\u5305\u62ec\u91dd\u5c0d\u8d85\u904e 1750 \u5104\u500b\u53c3\u6578\u7684\u6700\u5927\u6a21\u578b\uff0c\u9032\u884c\u5927\u578b\u8a9e\u8a00\u6a21\u578b\u8a13\u7df4\u548c\u63a8\u8ad6\u3002<\/p>\n\n\n\n<p>\u516b\u8def HGX H200 \u63d0\u4f9b\u8d85\u904e 32 petaflops \u7684 FP8 \u6df1\u5ea6\u5b78\u7fd2\u904b\u7b97\u548c 1.1TB \u7e3d\u9ad8\u983b\u5bec\u8a18\u61b6\u9ad4\uff0c\u53ef\u5728\u751f\u6210\u5f0f\u4eba\u5de5\u667a\u6167\u548c\u9ad8\u6548\u80fd\u904b\u7b97\u61c9\u7528\u4e2d\u5be6\u73fe\u6700\u9ad8\u6027\u80fd\u3002<\/p>\n\n\n\n<p>\u7576\u8207\u642d\u8f09\u8d85\u5feb\u901f NVLink-C2C \u4e92\u9023\u6280\u8853\u7684 NVIDIA Grace\u2122 CPU \u642d\u914d\u4f7f\u7528\u6642\uff0cH200 \u5275\u9020\u4e86\u642d\u8f09 HBM3e \u7684 GH200 Grace Hopper \u8d85\u7d1a\u6676\u7247\uff0c\u9019\u662f\u4e00\u500b\u5c08\u9580\u8a2d\u8a08\u4f86\u7528\u65bc\u5927\u898f\u6a21\u9ad8\u6548\u80fd\u904b\u7b97\u548c\u4eba\u5de5\u667a\u6167\u61c9\u7528\u7684\u6574\u5408\u6a21\u7d44\u3002<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><strong>\u4f7f\u7528<\/strong><strong>NVIDIA<\/strong><strong>\u5168\u7aef\u8edf\u9ad4\u52a0\u901f\u4eba\u5de5\u667a\u6167<\/strong><strong><\/strong><\/p>\n\n\n\n<p>NVIDIA \u7684\u52a0\u901f\u904b\u7b97\u5e73\u53f0\u5f97\u5230\u5f37\u5927\u7684\u8edf\u9ad4\u5de5\u5177\u652f\u63f4\uff0c\u9019\u4e9b\u5de5\u5177\u4f7f\u958b\u767c\u4eba\u54e1\u548c\u4f01\u696d\u80fd\u5920\u5efa\u69cb\u4e26\u52a0\u901f\u5f9e\u4eba\u5de5\u667a\u6167\u5230\u9ad8\u6548\u80fd\u904b\u7b97\u7684\u751f\u7522\u5c31\u7dd2\u578b\u61c9\u7528\u7a0b\u5f0f\u3002\u9019\u5176\u4e2d\u5305\u62ec <a href=\"https:\/\/www.nvidia.com\/zh-tw\/data-center\/products\/ai-enterprise\/\" target=\"_blank\" rel=\"noreferrer noopener\">NVIDIA AI Enterprise<\/a> \u8edf\u9ad4\u5957\u4ef6\uff0c\u7528\u65bc\u8a9e\u97f3\u3001\u63a8\u85a6\u7cfb\u7d71\u548c\u5927\u898f\u6a21\u63a8\u8ad6\u7b49\u5de5\u4f5c\u8ca0\u8f09\u3002<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><strong>\u4e0a\u5e02\u6642\u9593<\/strong><strong><\/strong><\/p>\n\n\n\n<p>\u5168\u7403\u7cfb\u7d71\u88fd\u9020\u5546\u548c\u96f2\u7aef\u670d\u52d9\u4f9b\u61c9\u5546\u5c07\u65bc 2024 \u5e74\u7b2c\u4e8c\u5b63\u958b\u59cb\u63d0\u4f9b NVIDIA H200 \u3002<\/p>\n\n\n\n<p><\/p>\n\n\n\n<p><\/p>\n\n\n\n<h4 class=\"wp-block-heading has-text-align-right has-very-light-gray-to-cyan-bluish-gray-gradient-background has-background\">\ud83d\udfe6<strong>\u73fe\u5728\u5c31\u52a0\u5165&nbsp;<a href=\"https:\/\/www.facebook.com\/profile.php?id=100086628162118\" target=\"_blank\" rel=\"noreferrer noopener\">ioioTIMES \u81c9\u66f8\u7c89\u7d72\u5718<\/a>&nbsp;\u66f4\u591a\u4e92\u52d5\u3001\u66f4\u591a\u597d\u5eb7\u650f\u62b5\u52a0!!<\/strong><br>\ud83d\udfe6<strong>\u6211\u5011\u6709<a href=\"https:\/\/today.line.me\/tw\/v2\/publisher\/103117\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">LINE TODAY<\/a>\u983b\u9053\u4e86\uff0c\u5feb\u4f86\u8ffd\u8e2a\u6211\u5011\u5427!!&#8211;\u6700\u65b0\u79d1\u6280\u65b0\u805e \u76e1\u5728\u4f60\u624b<\/strong><\/h4>\n","protected":false},"excerpt":{"rendered":"<p>NVIDIA (NASDAQ: NV<\/p>\n","protected":false},"author":3,"featured_media":49547,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"footnotes":""},"categories":[13],"tags":[7805,1287,71],"class_list":["post-49540","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","tag-hgx-h200","tag-hopper","tag-nvidia"],"_links":{"self":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/49540"}],"collection":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=49540"}],"version-history":[{"count":2,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/49540\/revisions"}],"predecessor-version":[{"id":49551,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/posts\/49540\/revisions\/49551"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=\/wp\/v2\/media\/49547"}],"wp:attachment":[{"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=49540"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=49540"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ioiotimes.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=49540"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}