Tried running an AI model on my gaming laptop, the fan noise was unbearable
I read all these posts about local LLMs being the way to go for privacy, so I fired up a 7B model on my RTX 3070 laptop. It worked, but the thing sounded like a jet engine taking off from my desk. After 20 minutes the room felt like a sauna and I couldn't hear my own thoughts. I shut it down and went back to using a cloud API. Guess my setup just isn't built for that kind of load. Anyone here run models locally without turning their rig into a space heater? What hardware are you using that makes it worth it?
Last year I sold my desktop with a 3080 partly because I got tired of the noise, so I get exactly where you're coming from. I figured buying a quieter laptop with a 4070 would fix it, but even a 13B model spins the fans up to full blast within minutes. The thing that changed my mind is realizing that cloud APIs are honestly fine for most of what I do, and the few times I need offline access, I just accept the noise and keep sessions short. My daily driver is now a small used Mac mini for the occasional local test, and everything else goes through the web.