Most people waste their GPU's potential, but I found a way to use every spare cycle

Published Aug 5, 2026, 7:00 AM EDT Maker, meme-r, and unabashed geek, Joe has been writing about technology since starting his career in 2018 at KnowTechie. He's covered everything from Apple to apps and crowdfunding and loves getting to the bottom of complicated topics. In that time, he's also written for SlashGear and numerous corporate clients before finding his home at XDA in the spring of 2023. He was the kid who took apart every toy to see how it worked, even if it didn't exactly go back together afterward. That's given him a solid background for explaining how complex systems work together, and he promises he's gotten better at the putting things back together stage since then. If you want to get the most value from your graphics card, you might use it for various productivity tasks when not gaming. That's one way to get more out of your flagship GPU if you're lucky enough to have one, but midrange GPUs are just as capable when you don't need lots of VRAM. But all of those require you to be actively using your PC. I wanted to fill in the downtime, the hours that I'm not using my GPU because I'm out of the house or asleep. The time that would otherwise be wasted capacity. Now, I'm not trying to run a data center at home and get 99.9% uptime. Just handle some tasks while I'm not using my PC for other things, and maybe reduce a few subscriptions from my list at the same time. A local LLM server Private ChatGPT, no subscription Whatever your feelings are about the ethics of cloud LLM providers, running local models is an effective use of your GPU resources. I've used a variety of backends, but now prefer using Lemonade on every machine, as I can build it once and then transfer those settings to any other hardware config. This is then served on my LAN to other devices via a web UI for chat and via an OpenAI-compatible endpoint for things like powering my coding agents. It's not used all the time, but it is available all the time, so that when I need it, it's already there. Lemonade Lemonade is an orchestrator for serving local and cloud LLMs for your needs. Local image generation with SwarmUI Midjourney at home, kind of Running a local image generation tool like SwarmUI keeps your data private. You don't need a particularly powerful GPU, but Nvidia graphics cards with 8GB or more of VRAM are best suited for this particular tool. It's not going to be as fast as online generators, but it runs on your hardware, doesn't cost a dime beyond your electricity bill, and you know your prompts aren't being used for training data. You also get a ton of powerful node-based tools to tinker with, and a Grid Generator that shows you how different parameter changes affect the generated images. And depending on the prompt's complexity and your GPU, this could be an overnight task. SwarmUI SwarmUI provides an image-generation server that runs on your hardware. A fully local voice assistant for Home Assistant Whisper hears, Piper answers Home Assistant is already powerful, but giving it access to an LLM running on another PC lets you set up a voice assistant pipeline that gives you local control over your smart home. All of this runs through the Home Assistant app on my smartphone, where it listens for a wake word before taking control of my smart home hub. It uses hardly any GPU cycles because short tasks don't take long to handle, but it's an important step toward reducing my reliance on the cloud. Voice assistants used to trade your privacy for convenience, but now locally hosted voice assistants are nearly as easy to use while keeping your voice inside your home. Whisper Whisper is a great way to transcribe audio for free on your machine, and it'll do so with high accuracy, too. It's a fantastic application with a lot of power, and you can set it up in minutes. Home Assistant OS Windows, macOS, Linux iOS compatible Yes Home Assistant gives you local control over your smart home. Frigate NVR object detection Cameras that actually understand Another thing I don't like having the cloud handle is the footage from my home security cameras. We've all seen the reports of misused footage from the companies entrusted with looking after it, and keeping it on my NAS is a smarter choice. And with Frigate paired with a local LLM, I don't miss out on motion detection. It can also send me summaries at the end of the day, and handle package detection without paying for Frigate+. Frigate Frigate is a self-hosted NVR app for your home security cameras. Jellyfin transcoding and AV1 library re-encoding Terabytes of space back on the NAS Handbrake will automatically select hardware-accelerated codecs whenever possible. In FFmpeg, the user will need to specify a codec that can be accelerated by the GPU.Handbrake also has batch editing, but with the right flags and arguments, FFmpeg can be much more powerful. With Jellyfin having native transcoding, you might think that it doesn't matter what form your media library is saved in. And for the most part, that's true enough. It'll transcode any formats your client can't play back, but there is one reason you might want to re-encode your library. That's space on your NAS. Hard drive prices are no joke at the moment, but even when they were cheaper, why wouldn't you reduce the size of your saved files instead of buying more storage? Re-encoding with Handbrake or other tools to AV1 will reclaim terabytes of space depending on how large your library is, and it's something that your GPU can do overnight while you sleep. Jellyfin iOS compatible Yes Android compatible Yes Jellyfin is one of the best Plex alternatives you can get, and that's thanks to its open-source nature and powerful set of features. There are apps for basically every platform and it's completely free to run your very own server. Handbrake Handbrake is free to download and use. Hunting pulsars with Einstein@Home Science with spare cycles If you don't need or want to run any of the other tools on this list, how about running some science? BOINC is a little piece of software that runs on your computer and lets you donate GPU cycles to various science projects. These include biomedical research, cognitive science, and my favorite, Einstein@Home, which searches for pulsars and gravitational waves. As of late 2024, the project has found 55 radio pulsars, 39 gamma-ray pulsars, and over 90 new neutron stars. That's a lot of science, found in large data sets with the numbers crunched by over 100K desktop and laptop PCs running the BOINC software. And it could be you that finds the next one. Einstein@Home Einstein@Home lets you participate in science. Space science. Isn't that cool? The only running cost is electricity It's no secret that GPUs are expensive hardware, even at the entry level. But once you've paid that tax, there are so many things you can do with them. Even when you're not actively using them for games or productivity tasks. With overnight scheduling or programs that are aware enough to know when your PC is otherwise idle, you can get even more value out of your GPU and maybe even contribute to science along the way.