I’m using a Crucial 48GB SODIMM stick in my FW12. Exact part number is CT48G56C46S5.M16B1.
I often run graphics intensive games and run LLM inference on this machine’s iGPU. The cooling of the actual CPU/GPU die is superb, no complaints there, it never goes above 85dC when connected to my 45W charger. But it appears that my RAM often stays in alarming temperature ranges. On basically idle it’s never below 60dC, while running games it often reaches 75-80dC, and in the peak of running LLM inference via llama.cpp through LM Studio it reaches a peak of 96dC. I have confirmed that there are no foam pads or weird obstacles near the RAM module. Is this something I need to be worried about or can I let it be?
The air intakes of the FL12 are really small and at the hinges, making the cooling difficult, so the solution is this? The RAM sits between the CPU and the battery, if these holes cool down the CPU by 15C and the battery by 6C, the RAM will be about 10.5C cooler
The speed holes mentioned in that thread would not cool the RAM in particular, because they’re not anywhere close to it. Moreover it would probably ruin the intentional airflow design the engineers came up with and silently cause some other components to overheat.
I noticed there is a foam pad below the RAM stick, which touches the actual RAM chips at all times. Could removing this bad be a good idea?
Is there any way to direct the air flow to the RAM? More intake (holes or not) on the right side of the chassis?
I have an FL13 it has stickers saying RAM channel 0/1 below the RAM stick, below the stickers the mainboard has some exposed metal dots on the PCB. I think the FL12’s foam serves the same purpose – prevents the RAM from conducting the mainboard. If you have non-conductive thermal pads I think you can remove the foam. However please discuss with other FL12 owners and find out whether it’s save to do so.
I have decided to eliminate the source of the problem and stop running LLMs on my computer for long amounts of time as a whole. Even during heavy graphics loads it does not go beyond about 83dC, and that will be sustained for an absolute max of 2-3 hrs a day, which sounds safe enough to me.
Would it be worth trying to connect it to the keyboard using thermal pads or something similar so that the heat can be dissipated better? That just came to mind. But then the RAM cover would be in the way, and I’m also not sure how it would generally affect the RAM, since it would be exposed to different stresses like pressure from the cover/keyboard.