--- title: AI At Home Part 1: A Box Of Scraps | daily.dev description: A software developer builds a homemade AI inference server using cheap e-waste AMD V620 workstation GPUs (originally overproduced for failed cloud-gaming... canonical: https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI At Home Part 1: A Box Of Scraps | daily.dev og:description: A software developer builds a homemade AI inference server using cheap e-waste AMD V620 workstation GPUs (originally overproduced for failed cloud-gaming... og:url: https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp og:image: https://api.daily.dev/og/posts/EBO2ACFNP.png og:image:alt: AI At Home Part 1: A Box Of Scraps og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # AI At Home Part 1: A Box Of Scraps **[Hacker News](https://daily.dev/sources/hn)** · 9 min read · 4 upvotes · 0 comments ## Summary A software developer builds a homemade AI inference server using cheap e-waste AMD V620 workstation GPUs (originally overproduced for failed cloud-gaming projects), a salvaged 2017 X299 motherboard, and off-brand Arduino fan controllers running custom PWM code. After wrestling with BIOS settings (Resizable BAR, MMIO High Size), no-fan server GPUs, and a deafening 10,000 RPM fan setup, the rig successfully boots Ubuntu 24.04 and runs llama.cpp with Gemma4 and Deepseek V4 Flash models. The build is part one of a series, with the next installment covering performance tuning. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is the AMD V620 GPU and why is it cheap on the used market? The AMD V620 is a workstation GPU repurposed for cloud gaming data centers, built from a workstation chip with extra RAM added and video outputs removed, never sold directly to consumers. It has 32GB of VRAM per card. AMD overproduced them before the cloud gaming market collapsed due to latency issues, leaving large surplus stock that e-waste resellers now sell cheaply on eBay, despite AMD's software support for AI workloads being notoriously weak. _Anyone piecing together budget GPU rigs for local LLM inference can compare notes on daily.dev._ ### Why won't an old X299 motherboard boot with four GPUs installed? Older X299 motherboards from 2017 were not designed to address the huge combined VRAM of four modern GPU cards, so they fail to boot until Resizable BAR and MMIO High Size are enabled. Both settings live deep in advanced BIOS menus and are separate options that must both be turned on for the board to recognize all cards and start up correctly. _Builders wiring up multi-GPU rigs on older boards can dig into BIOS quirks like this on daily.dev._ ### Why don't data center GPUs like the AMD V620 have cooling fans? Server-oriented GPUs like the V620 ship without fans because they are designed to rely on airflow from a wall of loud server-chassis blower fans rather than onboard cooling. For a home build with four cards stacked side by side, a custom 3D-printed shroud attaching an 80mm fan across two cards at a time solves the airflow problem better than the small, extremely loud 40mm shroud fans commonly sold online. _Anyone repurposing server GPUs at home can trade cooling hacks like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 44 comments across hackernews (as of 2026-09-13). **TL;DR:** Commenters largely validate that AMD GPUs (V620s and similar) now work well with ROCm and llama.cpp, but a big side debate questions whether building expensive local AI rigs makes economic sense versus just paying for cloud/API inference like DeepSeek. **Sentiment:** 45% positive · 45% mixed · 10% skeptical **The case for** - ROCm and AMD driver support has matured significantly and now works well out-of-the-box with llama.cpp. - Several people report similar successful builds using cheap secondhand AMD Instinct/Radeon cards. - Local hosting offers independence from cloud providers who can nerf, remove, or rug-pull access to models. - The build is praised as a fun, satisfying hardware tinkering project regardless of cost-effectiveness. **The pushback** - Multiple commenters argue the economics don't work out erence from providers like DeepSeek is far cheaper than the electricity and hardware cost of self-hosting. - Local models running on this hardware are much slower than renting equivalent compute, especially for larger models. - The V620 cards are hard to cool quietly, requiring loud high-RPM fans or DIY shroud/fan experimentation. - Local inference still depends on periodically updated models, so full independence from cloud/model providers isn't really achieved. **By community** - hackernews (mixed): Praise for how far AMD/ROCm has come is mixed with a long economic debate over whether local AI rigs are worth the cost versus cloud inference. **Hottest debate:** Whether building a local AI rig like this is a smart investment or a poor economic decision compared to just paying for cloud/API inference. **Open questions** - Will local self-hosted setups ever match the cost-effectiveness of cloud inference as hardware and electricity prices fluctuate? - Can local models be kept relevant long-term without access to continually updated frontier models? - What case/chassis options exist for fitting four dual-slot GPUs, since most ATX cases don't support this? **Highlights** > Ah, the price has gone up on the AI Pro 9700, again, I see. So, yeah, it's a worse deal now. But, the V620 is not really a good deal, either, at $450 on eBay. It's roughly half the performance of the newer card. Usable but not blazing. And, you probably don't want to deal with cooling them; if you already have a 3D printer, you can print the shrouds for a few bucks and buy the fans for a few more bucks. But, if you don't have a 3D printer, the shrouds you can buy on eBay are for the tiny fans that have to run at ludicrous speed and have to be so fucking loud it should be illegal. I went through three different shroud and fan combos before finding a set that worked in my case with this cards (about $100 worth of experimentation, probably). I should say, though, that I actually don't recommend buying anything right now. I wrote up my setup and made recommendations (and the main recommendation was "don't"). https://swelljoe.com/post/how-i-run-local-llms/ The V620 at $450 or the Radeon AI Pro 9700 at ~$1400 are a good deal compared to everything else right now, but buying tokens from DeepSeek is a better deal. DeepSeek V4 Flash 0731 is better than anything you can host locally, they'll serve it to you at blistering fast speeds for pennies a day, and with 1 million token context. You could host a 2-bit quantization of it on a Strix Halo or four of these V620s, and it would run at a crawl on either one. The Strix Halo gets 9-13 t/s. Four V620s would, I guess, get two-three times that. Which is still too slow for comfortable interactive agentic use, and much slower than getting it from DeepSeek, and at two bits there is measurable loss. You're paying a lot more for self-hosted and you're getting worse models. Also, if you want four cards, you need a server-class motherboard and CPU and RAM. More money. It's all just a bad investment. Self-hosting is a bad idea if you don't already have the hardware, unless and until memory and GPU prices come down. Even at the prices I paid (before RAMpocalypse really kicked off; $2k for the Strix Halo, ~$350 for each V620), I wouldn't recommend it if you don't have a strong urge to tinker with hardware and it'll probably never pay for itself vs. buying inference from DeepSeek directly. Edit: The same shrouds for all the old Instinct cards also work for the V620, as they have the same dimensions and screw and cable layout. I tried several and ended up with this one: https://www.thingiverse.com/thing:7296707 > — [SwellJoe on hackernews · 2 comments](https://news.ycombinator.com/item?id=49291020) > Sure but the GPUs in the article are from 2021 and cost ~500-600$ a pop. The OP bought 4 so that's 2k, let's say another 500$ for the rest of the chassis and you stand at $2.5k for a monster server that will burn electricity and warm your house. Compare that to a $5k DGX than consumes much less and has the same amount of VRAM and there is a real question as whether this is worth doing at all (well aside from the cool factor). > — [belval on hackernews · 1 comments](https://news.ycombinator.com/item?id=49290855) > I mean, this is cool and all, but 2 DGX Sparks is like $10k, right? That's a huge upfront cost to run DS4 Flash 0731 which is definitely not comparable to higher-end Claude models. It's extra egregious when you consider how much it'd cost to run DS4 Flash 0731 via an API (how long would it take to run up a $10k bill doing what you're doing?). >and from someone that has been using Claude for a long time, I can definitely say I don't need it anymore. Not for the stuff I'm doing. I'm curious what this implies. It suggests that you were using Claude prior to this setup, so it doesn't seem like you were limited by security or local features? I can understand that someone not wanting to send their data to Anthropic (etc.) might be willing to pay for this kind of setup to accomplish that, but what was your motivation for doing this? If it matters, if I had enough money that I could blow $10k on 2 DGX Sparks without it being a significant cost I probably would for the hell of it, so I'm not digging on you if this is ultimately what this comes down to. But, as an investment, this doesn't seem like a very good deal. > — [nater5000 on hackernews · 5 comments](https://news.ycombinator.com/item?id=49290320) > "used to" in your sentence already answers your curiosity. ROCm runs most things just fine without any ceremony or difficulty. I have a couple of the same cards as covered in the article, bought before they got expensive (I'd recommend current Radeon AI Pro 9700 cards over the V620 now, though, as they have increased in price since I bought mine), and they're at the long end of the supported chart for ROCm...nearly EOL. But, they currently work great with llama.cpp and current ROCm 7.14. They're pretty fast and stable. I also have a Strix Halo 128GB, and while the Strix Halo can run bigger models, the dedicated GPUs are quite a bit faster. And, since the best models you can run at home are probably Qwen 3.6 27B and Gemma 4 31B, and those both fit comfortably on dual 32GB GPUs, I find I use the desktop more often than the Strix Halo. Anyway, there's very little reason to spend 3x or more for the Nvidia ecosystem these days. The only exception is the Strix Halo vs the Nvidia GB10 platform. Strix Halo was a no-brainer when it was half the price, but it's risen in price to be almost the same price as an Asus GX10. In that one instance, I think the Nvidia based Asus is a better choice. The Strix Halo is almost too slow to make use of 128GB. Some MoE models are comfortable, like Laguna S2.1, so it's probable that there will someday be an MoE model that is better then Qwen 3.6 27B or Gemma 4 31B that won't run on 64GB but will run on 128GB. I don't know of one, yet, though. AMD also used to focus almost entirely on their data center line and not pay any attention to their lower end stuff for more advanced AI features, but that's been changing, and ROCm support has broadened to include almost everything AMD ships now, including the smaller embedded stuff. I think they've realized that as long as the only way to develop AI for AMD was to have a quarter million dollars worth of hardware, they would always fall behind a platform that can be developed for and tested on consumer hardware. > — [SwellJoe on hackernews · 1 comments](https://news.ycombinator.com/item?id=49290310) > > The cards had some kind of metal cable guide fin on the back That's to slot into the front PCI brackets for full length cards. There's "full" size for PCI cards and most GPUs are considered "half" length cards. This is separate to Low Profile, so there are four possible slot compatibility configurations, namely HHHL, HHFL, FHFL, and FHHL. Real workstations(including Mac Pro) has slots cut in the case or a bracket to accept the card or that metal piece(which means a standardized solution to GPU sagging had existed even before PCIe was created). Also, on fans, it's not like blower fans are quieter at all, but in case anyone encounters a situation where turning up fans seem to be only creating noises without moving air or cooling yhe cards, it might be worth remembering that radial fans have higher static pressure and are better at forcing air through. edit: ps: people recreating this might want to know what case this is. The vast majority of ATX cases, even unnecessarily big ones, only has 7 PCI slots. Cases that can take four dual-slot cards is rare. > — [numpad0 on hackernews](https://news.ycombinator.com/item?id=49291758) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=49288293) · 78 points · 44 comments ## Similar posts on daily.dev - [Hands-On with the AMD Ryzen AI Halo](https://daily.dev/posts/hands-on-with-the-amd-ryzen-ai-halo-g0llzrz16) · Hacker News · 1 upvotes · 0 comments - [I quantized a local LLM on my home server and ditched cloud AI for smart home control entirely](https://daily.dev/posts/i-quantized-a-local-llm-on-my-home-server-and-ditched-cloud-ai-for-smart-home-control-entirely-sv0xgvzfn) · XDA Developers · 2 upvotes · 0 comments - [Untitled](https://daily.dev/posts/untitled-crxvfgirl) · SitePoint · 0 upvotes · 0 comments --- Tags: [#gpu](https://daily.dev/tags/gpu), [#self-hosting](https://daily.dev/tags/self-hosting), [#local-ai](https://daily.dev/tags/local-ai), [#homelab](https://daily.dev/tags/homelab), [#llama-cpp](https://daily.dev/tags/llama-cpp) [View this post on daily.dev](https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"AI At Home Part 1: A Box Of Scraps","url":"https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp"},"datePublished":"2026-08-13T19:01:26.848Z","dateModified":"2026-09-13T18:55:40.668Z","description":"A software developer builds a homemade AI inference server using cheap e-waste AMD V620 workstation GPUs (originally overproduced for failed cloud-gaming...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3af05b0584aba8c050026728331b0541?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/3af05b0584aba8c050026728331b0541?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"gpu,self-hosting,local-ai,homelab,llama-cpp","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"AI At Home Part 1: A Box Of Scraps"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-at-home-part-1-a-box-of-scraps-ebo2acfnp#faq","mainEntity":[{"@type":"Question","name":"What is the AMD V620 GPU and why is it cheap on the used market?","acceptedAnswer":{"@type":"Answer","text":"The AMD V620 is a workstation GPU repurposed for cloud gaming data centers, built from a workstation chip with extra RAM added and video outputs removed, never sold directly to consumers. It has 32GB of VRAM per card. AMD overproduced them before the cloud gaming market collapsed due to latency issues, leaving large surplus stock that e-waste resellers now sell cheaply on eBay, despite AMD's software support for AI workloads being notoriously weak. Anyone piecing together budget GPU rigs for local LLM inference can compare notes on daily.dev."}},{"@type":"Question","name":"Why won't an old X299 motherboard boot with four GPUs installed?","acceptedAnswer":{"@type":"Answer","text":"Older X299 motherboards from 2017 were not designed to address the huge combined VRAM of four modern GPU cards, so they fail to boot until Resizable BAR and MMIO High Size are enabled. Both settings live deep in advanced BIOS menus and are separate options that must both be turned on for the board to recognize all cards and start up correctly. Builders wiring up multi-GPU rigs on older boards can dig into BIOS quirks like this on daily.dev."}},{"@type":"Question","name":"Why don't data center GPUs like the AMD V620 have cooling fans?","acceptedAnswer":{"@type":"Answer","text":"Server-oriented GPUs like the V620 ship without fans because they are designed to rely on airflow from a wall of loud server-chassis blower fans rather than onboard cooling. For a home build with four cards stacked side by side, a custom 3D-printed shroud attaching an 80mm fan across two cards at a time solves the airflow problem better than the small, extremely loud 40mm shroud fans commonly sold online. Anyone repurposing server GPUs at home can trade cooling hacks like this on daily.dev."}}]} ```