{"id":58938,"date":"2022-03-18T15:02:36","date_gmt":"2022-03-18T13:02:36","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=58938"},"modified":"2025-09-04T19:21:30","modified_gmt":"2025-09-04T16:21:30","slug":"google-applies-deep-learning-to-ai-chip-development","status":"publish","type":"post","link":"https:\/\/u1f987.com\/en\/google-applies-deep-learning-to-ai-chip-development\/","title":{"rendered":"Google Applies Deep Learning to AI-Chip Development"},"content":{"rendered":"<p>Google and the University of California, Berkeley have created PRIME, a deep-learning algorithm that helps design fast, compact processors for artificial intelligence tasks.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">Presenting PRIME, a data-driven approach for architecting hardware accelerators that trains a <a href=\"https:\/\/twitter.com\/hashtag\/DeepLearning?src=hash&#038;ref_src=twsrc%5Etfw\">#DeepLearning<\/a> model on existing accelerator data, improves runtime and chip area usage by 1.2 \u2014 1.5X, and can generate accelerators for unseen applications \u2192 <a href=\"https:\/\/t.co\/E0PcQMg3d4\">https:\/\/t.co\/E0PcQMg3d4<\/a> <a href=\"https:\/\/t.co\/NdQWQgZ4AA\">pic.twitter.com\/NdQWQgZ4AA<\/a><\/p>\n<p>\u2014 Google AI (@GoogleAI) <a href=\"https:\/\/twitter.com\/GoogleAI\/status\/1504524701762269191?ref_src=twsrc%5Etfw\">March 17, 2022<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<p>The new approach designs AI-chip architectures from existing designs and performance metrics.<\/p>\n<p>The team said PRIME-built chips have latency up to 50% lower than those produced by classical approaches. The deep-learning approach also cut the time to generate designs by up to 99%.<\/p>\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1251\" height=\"534\" src=\"https:\/\/u1f987.com\/wp-content\/uploads\/google-prime-dl-1.gif\" alt=\"Google applies deep learning to AI-chip development\" class=\"wp-image-167986\"\/><figcaption>Operation of the PRIME algorithm. Data: Google.<\/figcaption><\/figure>\n<p>The researchers compared the performance of PRIME-built chips with EdgeTPU accelerators across nine AI applications, including image-classification models MobileNetV2 and MobileNetEdge. They emphasised that the designs were optimised for each application.<\/p>\n<p>The PRIME approach improved latency by a factor of 2.7 and reduced die area by a factor of 1.5. This could lower chip costs and reduce power consumption, the researchers said.<\/p>\n<p>In addition, the AI-designed chips outperformed in all nine applications tested. Only three applications showed higher latency than the designs created via modelling.<\/p>\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"400\" height=\"400\" src=\"https:\/\/u1f987.com\/wp-content\/uploads\/google-prime-dl-2.png\" alt=\"Google applies deep learning to AI chip development\" class=\"wp-image-167987\" srcset=\"https:\/\/u1f987.com\/wp-content\/uploads\/google-prime-dl-2.png 400w, https:\/\/u1f987.com\/wp-content\/uploads\/google-prime-dl-2-300x300.png 300w, https:\/\/u1f987.com\/wp-content\/uploads\/google-prime-dl-2-150x150.png 150w\" sizes=\"auto, (max-width: 400px) 100vw, 400px\" \/><figcaption>Comparison of latency across nine applications (lower is better). Data: Google.<\/figcaption><\/figure>\n<p>According to the researchers, PRIME holds promising prospects, including creating microchips for applications requiring solutions to complex optimisation problems, and using designs of low-performance chips as training data.<\/p>\n<p>In June 2021, Google described <a href=\"https:\/\/u1f987.com\/en\/news\/google-cuts-chip-design-time-from-months-to-six-hours-with-ai\">using reinforcement learning to speed up chip design<\/a> from several months to six hours.<\/p>\n<p>In October, the company unveiled the Pixel 6 and Pixel 6 Pro <a href=\"https:\/\/u1f987.com\/en\/news\/google-unveils-pixel-6-and-pixel-6-pro-with-tensor-processor-for-machine-learning\">with a Tensor chip for machine learning<\/a> of its own design.<\/p>\n<p>In August, Samsung began using artificial intelligence to <a href=\"https:\/\/u1f987.com\/en\/news\/samsung-deploys-ai-to-design-mobile-processors\">automate the process of designing computer chips<\/a>.<\/p>\n<p>Subscribe to ForkLog AI on Telegram: <a href=\"https:\/\/t.me\/forklogAI\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">ForkLog AI<\/a> \u2014 all the news from the world of AI!<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Google and the University of California, Berkeley have developed PRIME, a deep-learning algorithm that helps design fast, compact processors for AI tasks.<\/p>\n","protected":false},"author":1,"featured_media":58939,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"1","news_style_id":"1","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,738],"class_list":["post-58938","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-google"],"aioseo_notices":[],"amp_enabled":true,"views":"17","promo_type":"1","layout_type":"1","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/58938","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/comments?post=58938"}],"version-history":[{"count":1,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/58938\/revisions"}],"predecessor-version":[{"id":58940,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/58938\/revisions\/58940"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media\/58939"}],"wp:attachment":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media?parent=58938"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/categories?post=58938"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/tags?post=58938"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}