Assuming you’re in the northern hemisphere: For this winter it’s fine. It’ll gently heat your home while you game like it’s 1999. No worries 😁
However, once it starts to warm up you’ll want to send that motherboard+RAM+CPU to your local HAZMAT trash pickup/facility and get something newer. Might I suggest a nice 2020-ish desktop CPU? With a motherboard that supports Coreboot, of course!
…and get yourself a nice Nvidia (sadly, because AMD and Intel are still far behind) GPU with at least 12GB of VRAM so you can have fun with the open source AI stuff (it’s a blast!). The more VRAM the better though so if you can pick up a 4060 Ti with 16GB cheap this spring that’ll be your best budget buy (endless uncensored fun) 👍
Seriously: If you haven’t got the hardware to run Stable Diffusion locally you’re missing out! It’s as fun and addicting as a really good game. Running it on some cloud service isn’t the same because at best they’ll be running stuff that’s weeks or months out of date (which is like a million years in AI time) and they don’t give you the same level of control/possibilities that you get when running your own stuff locally (run whatever models/LoRAs you want, whatever extensions you want, generating images without having to worry about overbearing censorship because it is that bad on public AI services–paid or not!).
Oh I’m already using AUTOMATIC1111 on this computer with my RTX 2080 SUPER (using Debian GNU/Linux). It runs really good on it actually, I’m only using 8GB of VRAM too, It takes like 5-7 seconds to generate images, truly amazing stuff. I may be able to push out a video today or tomorrow, the computer can run just fine even with this current RAM issue.
I was just playing GTA V like 10 minutes ago and it ran well like 50-70FPS (sometimes drops to 40-45FPS depening on the area) on max settings 1440p, barely any stutters. I can run basically everything I need on it with Libreboot now. I’m selling off my main motherboard right now.
The only thing I can’t seem to get running is local LLMs because my Xeon processor is AVX-only and not AVX2. I was trying to use GPT4All, I was able to use the UI and try to download a model, but aftering it finished downloading it crashes and I can’t open it again. As far as I looked into it, its just the current limitation with my processor.
In the future I may start looking into using Dasharo firmware with the MSI Z690-A DDR5 since I want to use new models like Mixtral 8x7B, I heard it was really good. The older motherboard I’m using here is the Dell Precision T1650 motherboard from 2012, its only 32GB of RAM max, and newer models like Mixtral need up to 64GB of RAM.
I couldn’t check the exact score of how much open-source code there is on here, I did have to inject blobs / vendor firmware into the libreboot.rom file. It should probably be more open than newer motherboards, considering its older, along with Libreboots blob reduction policy.
The only small problem I’m facing right now is for some reason I have to leave this computer on for like an hour to make my games run better. I open Steam and I guess just let the Vulkan shaders process in the background.
Also, I tried running NVIDIAs open source drivers and it didn’t work for me. I couldn’t passthrough my GPU to my VM (using KVM hypervisor) either. Maybe, this is just due to me using NVIDIAs full proprietary drivers or maybe I just suck lol I would like more people to try this out though.
Hopefully one day we can achieve even more freedom with better hardware, maybe with hardware like RISC-V =)
Assuming you’re in the northern hemisphere: For this winter it’s fine. It’ll gently heat your home while you game like it’s 1999. No worries 😁
However, once it starts to warm up you’ll want to send that motherboard+RAM+CPU to your local HAZMAT trash pickup/facility and get something newer. Might I suggest a nice 2020-ish desktop CPU? With a motherboard that supports Coreboot, of course!
https://doc.coreboot.org/mainboard/index.html
…and get yourself a nice Nvidia (sadly, because AMD and Intel are still far behind) GPU with at least 12GB of VRAM so you can have fun with the open source AI stuff (it’s a blast!). The more VRAM the better though so if you can pick up a 4060 Ti with 16GB cheap this spring that’ll be your best budget buy (endless uncensored fun) 👍
Seriously: If you haven’t got the hardware to run Stable Diffusion locally you’re missing out! It’s as fun and addicting as a really good game. Running it on some cloud service isn’t the same because at best they’ll be running stuff that’s weeks or months out of date (which is like a million years in AI time) and they don’t give you the same level of control/possibilities that you get when running your own stuff locally (run whatever models/LoRAs you want, whatever extensions you want, generating images without having to worry about overbearing censorship because it is that bad on public AI services–paid or not!).
Oh I’m already using AUTOMATIC1111 on this computer with my RTX 2080 SUPER (using Debian GNU/Linux). It runs really good on it actually, I’m only using 8GB of VRAM too, It takes like 5-7 seconds to generate images, truly amazing stuff. I may be able to push out a video today or tomorrow, the computer can run just fine even with this current RAM issue.
I was just playing GTA V like 10 minutes ago and it ran well like 50-70FPS (sometimes drops to 40-45FPS depening on the area) on max settings 1440p, barely any stutters. I can run basically everything I need on it with Libreboot now. I’m selling off my main motherboard right now.
The only thing I can’t seem to get running is local LLMs because my Xeon processor is AVX-only and not AVX2. I was trying to use GPT4All, I was able to use the UI and try to download a model, but aftering it finished downloading it crashes and I can’t open it again. As far as I looked into it, its just the current limitation with my processor.
In the future I may start looking into using Dasharo firmware with the MSI Z690-A DDR5 since I want to use new models like Mixtral 8x7B, I heard it was really good. The older motherboard I’m using here is the Dell Precision T1650 motherboard from 2012, its only 32GB of RAM max, and newer models like Mixtral need up to 64GB of RAM.
I couldn’t check the exact score of how much open-source code there is on here, I did have to inject blobs / vendor firmware into the libreboot.rom file. It should probably be more open than newer motherboards, considering its older, along with Libreboots blob reduction policy.
The only small problem I’m facing right now is for some reason I have to leave this computer on for like an hour to make my games run better. I open Steam and I guess just let the Vulkan shaders process in the background.
Also, I tried running NVIDIAs open source drivers and it didn’t work for me. I couldn’t passthrough my GPU to my VM (using KVM hypervisor) either. Maybe, this is just due to me using NVIDIAs full proprietary drivers or maybe I just suck lol I would like more people to try this out though.
Hopefully one day we can achieve even more freedom with better hardware, maybe with hardware like RISC-V =)