I spent 3800$ on a new PC to use AI... Do I regret it?

My old GTX 1080Ti was falling behind on AI. That's why I bought a new RTX 4090, that needed a fully new PC. Will the more than 3000 dollars spent be worth the cost? Well, let's live test and react to it. Sorry for the off-topic video, and with a poorly set up camera xD. Hope it helps someone out there but it mainly serves as an anouncement. Now my pc won't be a set back anymore, so expect more videos :3 Join our discrod server: https://discord.gg/8TSMNHvGCH to learn about this and more! Thanks to everyone that helped with this video! (Lefourbe did a lot more testing that I didn't know where to add. Ask on discord if you are interested) ------------- Links used in the VIDEO ---------- GTX 1080 TI vs RTX 4090: https://technical.city/es/video/GeForce-GTX-1080-Ti-vs-GeForce-RTX-4090 -----Optimizing your pc: --DarkStoorm's gide (check pinned comment too): https://gist.github.com/DarkStoorM/159d723c93dde06b17b6f251c46cd827 --Kenjin's tips: If you have not so much VRAM and want to optimize your upscaling without using --medvram or --lowvram You can check Tiled diffusion with Controlnet Tiles to optimize furthermore, it's a very good tool. You'll need to find the best parameters for you though to get the best speed and quality: https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111. Removing --medvram or --lowvram if you have enough memory and they are active. If you have --no-half, you can replace it by --upcast-sampling (see https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/8782) Adding set PYTORCH_CUDA_ALLOC_CONF= garbage_collection_threshold:0.9,max_split_size_mb:512 in webui-user.bat to reduce memory fragmentation if you run the application for a long period of time. His Sources: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations --Another option you have is just installing ComfyUI, as it is more optimized and runs SDXL better than A1111. It's a bit more complex tho: https://github.com/comfyan