{"id":13827,"date":"2026-02-03T08:13:39","date_gmt":"2026-02-03T08:13:39","guid":{"rendered":"https:\/\/davelevy.info\/wiki\/?p=13827"},"modified":"2026-02-03T08:45:55","modified_gmt":"2026-02-03T08:45:55","slug":"ai-vs-hpc-whats-the-difference","status":"publish","type":"post","link":"https:\/\/davelevy.info\/wiki\/ai-vs-hpc-whats-the-difference\/","title":{"rendered":"AI vs HPC, what&#8217;s the difference."},"content":{"rendered":"\n<p>What is the architecture of an AI Cluster and how does it differ from other  distributed computing architectures? These are my notes, a work in progress. <\/p>\n\n\n\n<!--more-->\n\n\n\n<p> Schwartz (or someone that advised him) articulated that there were three super-architectures, Virtualisation Hosts, Web Farm servers, and Super-computers. In <a href=\"https:\/\/www.linkedin.com\/pulse\/strategy-alignment-david-levy\/\">my article on IS Strategy<\/a>, on LinkedIn, (April 2020) I described them as <\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>Today I describe the systems as business systems, community systems and High Performance Computing.<\/p>\n<\/blockquote>\n\n\n\n<p>I hadn&#8217;t realised that it was that old. It would seem that now I have to add AI clusters as 4th super-architecture. <\/p>\n\n\n\n<p>I <a href=\"https:\/\/chatgpt.com\/c\/697480b9-4124-8332-aa33-88e2a9b3441b\">asked ChatGPT<\/a>, who now it seems serve AI with ads, who said<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>A <strong>supercomputer<\/strong> is engineered to solve <em>physics equations about reality with extreme precision<\/em>, while an <strong>AI cluster node<\/strong> is engineered to <em>multiply giant matrices as fast as physically possible<\/em> to train neural networks.<\/p>\n<\/blockquote>\n\n\n\n<p>And<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li>&#8220;AI doesn\u2019t usually need extreme numerical precision \u2014 it needs <strong>throughput<\/strong>.&#8221;<\/li>\n\n\n\n<li>They use different interconnects<\/li>\n\n\n\n<li>Power and power density are different, AI systems draw more power, with a higher density and thus require more expensive cooling.<\/li>\n\n\n\n<li>Different RAM and Chips<\/li>\n<\/ol>\n\n\n\n<p>I looked for white papers, and found these<\/p>\n\n\n\n<ol class=\"wp-block-list\">\n<li><a href=\"https:\/\/insideainews.com\/2019\/11\/27\/supercomputers-and-machine-learning-a-perfect-match\/\">https:\/\/insideainews.com\/2019\/11\/27\/supercomputers-and-machine-learning-a-perfect-match\/<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.ibm.com\/think\/topics\/hpc\">https:\/\/www.ibm.com\/think\/topics\/hpc<\/a><\/li>\n\n\n\n<li><a href=\"https:\/\/www.glennklockwood.com\/garden\/HPC-vs-AI\">https:\/\/www.glennklockwood.com\/garden\/HPC-vs-AI<\/a>, Glenn works for VASTdata, previously at Microsoft.<\/li>\n<\/ol>\n\n\n\n<p>In order to understand the software models, I asked gemini, , and it <a href=\"https:\/\/share.google\/aimode\/ubYl0YHyHpjNn6HS6\">answered with this<\/a>. More to read.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>What is the architecture of an AI Cluster and how does it differ from other distributed computing architectures? These are my notes, a work in progress.<\/p>\n","protected":false},"author":1,"featured_media":13828,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_memberships_contains_paid_content":false,"footnotes":"","_share_on_mastodon":"0"},"categories":[3],"tags":[2245,2011,622,1200,1889,911],"class_list":["post-13827","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-architecture","tag-artificialintelligence","tag-distributed-systems","tag-hpc","tag-supercomputers","tag-technology"],"share_on_mastodon":{"url":"https:\/\/mastodon.social\/@davelevy_eu\/116005694428716249","error":""},"jetpack_featured_media_url":"https:\/\/davelevy.info\/wiki\/wp-content\/uploads\/2026\/02\/computerhall-ismailenes-ayhan-lVZjvw-u9V8-unsplash-w1080.jpg","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/posts\/13827","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/comments?post=13827"}],"version-history":[{"count":3,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/posts\/13827\/revisions"}],"predecessor-version":[{"id":13831,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/posts\/13827\/revisions\/13831"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/media\/13828"}],"wp:attachment":[{"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/media?parent=13827"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/categories?post=13827"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/davelevy.info\/wiki\/wp-json\/wp\/v2\/tags?post=13827"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}