{"id":622440,"date":"2026-10-06T16:32:09","date_gmt":"2026-10-06T16:32:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/622440\/"},"modified":"2026-10-06T16:32:09","modified_gmt":"2026-10-06T16:32:09","slug":"someone-turned-their-phone-into-a-working-external-gpu","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/622440\/","title":{"rendered":"Someone turned their phone into a working external GPU"},"content":{"rendered":"<p>You don&#8217;t have to buy a <a href=\"https:\/\/www.howtogeek.com\/you-dont-need-an-egpu-anymoreheres-why-gaming-laptops-finally-caught-up\/\" target=\"_blank\" rel=\"nofollow noopener\">pricey video card enclosure<\/a> to add external graphics to your PC \u2014 the phone in your pocket might do. Reddit user StayLameBro has found a way to use his <a href=\"https:\/\/www.howtogeek.com\/iphone-17-pro-review\/\" target=\"_blank\" rel=\"nofollow noopener\">iPhone 17 Pro Max<\/a> as an external GPU that assists his <a href=\"https:\/\/www.howtogeek.com\/apple-macbook-pro-m4-review\/\" target=\"_blank\" rel=\"nofollow noopener\">MacBook Pro M4<\/a>.<\/p>\n<p>The <a href=\"https:\/\/www.reddit.com\/r\/LocalLLaMA\/comments\/1wvz1ex\/i_made_my_iphone_a_second_gpu_for_my_24_gb\/\" rel=\"noopener noreferrer nofollow\" target=\"_blank\">new project<\/a> splits AI model tasks (in this case, Qwen 3.8 27B) between the Mac and iPhone. The laptop normally handles the first 40 layers of each AI token batch, while the iPhone&#8217;s GPU takes care of the rest. If there are more than 64,000 tokens involved, the Mac runs all the layers while the iPhone focuses on old key value pages (identifying labels and their matching data).<\/p>\n<p>The Mac also allocates 8-bit context (the complete set of background info, data, and memory) based on the iPhone&#8217;s free RAM, keeping the computer&#8217;s memory usage in check. On top of this, the handset contributes to prefills (processing all the inputs in a token prompt) beyond 512 tokens.<\/p>\n<p>Yes, the speed improvements are noticeable. When prefilling a 2,000-token file into an existing agent session, the MacBook and iPhone combination is between 29% to 44% faster than the Mac by itself. With a new 27,000-token agent session, the gains range between 36% for the project&#8217;s forked llama.cpp code and 46% versus the stock software.<\/p>\n<p>The code, setup process, and benchmarking scripts are available now <a href=\"https:\/\/github.com\/StayLameBro\/backburner\" rel=\"noopener noreferrer nofollow\" target=\"_blank\">in a GitHub repository<\/a>. You don&#8217;t need more than a standard 10Gbps USB-C cable to tether to the Mac, although you need at least an iPhone 15 Pro. An <a href=\"https:\/\/www.howtogeek.com\/iphone-18-pro-and-pro-max-have-a-camera-upgrade-dslr-users-will-actually-love\/\" target=\"_blank\" rel=\"nofollow noopener\">iPhone 18 Pro<\/a> should have &#8220;a lot more&#8221; headroom, the creator says.<\/p>\n<p>                        What problems would I have using my phone as an external GPU?<\/p>\n<p>            AI models and write speeds are both concerns<\/p>\n<p>The iPhone as an external GPU doesn&#8217;t solve every issue. StayLameBro notes that the phone doesn&#8217;t improve write performance below 64,000 tokens, leaving that task to the computer.<\/p>\n<p>                                                                                                                            <a href=\"https:\/\/www.howtogeek.com\/how-i-run-heavy-open-source-llms-for-free-without-a-gpu\/\" rel=\"nofollow noopener\" target=\"_blank\"><br \/>\n                        <img width=\"440\" height=\"248\" loading=\"lazy\" decoding=\"async\" alt=\"The chat window in Jan.ai.\" data-img-url=\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/10\/the-chat-window-in-jan-ai.png\" src=\"https:\/\/www.newsbeep.com\/nz\/wp-content\/uploads\/2026\/10\/the-chat-window-in-jan-ai.png\"\/><\/p>\n<p>                    <\/a><\/p>\n<p>                    Related<\/p>\n<p>\t\t<a href=\"https:\/\/www.howtogeek.com\/how-i-run-heavy-open-source-llms-for-free-without-a-gpu\/\" title=\"How I run heavy open-source LLMs for free without a GPU\" target=\"_blank\" rel=\"nofollow noopener\"><br \/>\n\t\t\tHow I run heavy open-source LLMs for free without a GPU<br \/>\n\t\t<\/a><\/p>\n<p class=\"display-card-excerpt\">You don&#8217;t need a fancy GPU to train and test your favorite large language models.<\/p>\n<p>He adds that he hasn&#8217;t benchmarked AI models with newer features, including Qwen 3.8-Flash-Next and DeepSeek V4.1-Flash. There&#8217;s no guarantee the iPhone will be as useful. There&#8217;s no mention of Android or Windows, either, making the existing version an all-Apple solution.<\/p>\n<p>Even so, the effort shows the potential for phones as computing aids. The improvements won&#8217;t compare to those of a <a href=\"https:\/\/www.howtogeek.com\/best-gpus-that-launched-in-2025-and-one-complete-flop\/\" target=\"_blank\" rel=\"nofollow noopener\">good video card<\/a>, but they could still save valuable time when you need to complete AI processes quickly.<\/p>\n","protected":false},"excerpt":{"rendered":"You don&#8217;t have to buy a pricey video card enclosure to add external graphics to your PC \u2014&hellip;\n","protected":false},"author":2,"featured_media":622441,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[342,111,139,69,145],"class_list":["post-622440","post","type-post","status-publish","format-standard","has-post-thumbnail","category-mobile","tag-mobile","tag-new-zealand","tag-newzealand","tag-nz","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/622440","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=622440"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/622440\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/622441"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=622440"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=622440"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=622440"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}