Hi, I have three Comet devices, two deployed, and I've noticed that they tend to run very warm when idle and will overheat with sufficient use.
Background:
I normally use my devices in P2P mode on my local network and run the graphics at "Ultra High".
I do have Tailscale installed on each, but again, I don't often use that capability.
My current workaround it to adhere (with thermal pads) 40mm heat sinks and fans to the hottest areas, which seem to the the top of the device on the logo, roughly the "upper" half of the device nearest the power input.
Before I applied the workaround, they would both idle at 65C. With the workaround, they idle at about 34C. For comparison, I am also running a PiKVM and it idles at about 34C at the hottest point of its case.
First, I'm wondering if others find their devices idle warm, and about what temperature they hit if the device stops working under load.
Second, I'm wondering if anyone knows of a console or other method for on-board temperature measurement of the device for logging and troubleshooting.
Third, I'd like to put in a feature request for the client software to report any onboard measurements if possible.
Thanks for reading and any feedback.
Hi, you can check this command in terminal cat /sys/class/thermal/thermal_zone0/temp.
Thanks for your suggestion! That answers one question.
Your post intrigued me so I opened up the second comet I received.
It would seem the rockchip SOC is not touching the heatsink. I had to add ~3mm of heat transfer pads for it to make contact:
I am not deploying this one yet, but if you have access to heat transfer tape, open yours up and test!
To check your temperatures before and after:
CPU: cat /sys/class/thermal/thermal_zone0/temp
NPU: cat /sys/class/thermal/thermal_zone1/temp
From what I can tell, throttling occurs at ~85C, thermal shutdown at 105C.
I will do some testing if time permit later on!
I just upgraded the firmware from 1.1->1.3.1->1.4.2 on this one, it's idling at 40C with the thermal pads (nothing is connected to it):
bash-4.4# cat /sys/class/thermal/thermal_zone0/temp
40000
bash-4.4# cat /sys/class/thermal/thermal_zone1/temp
39900
thx for the tip. i did the same. to test it out, i ran youtube playing with audio via KVM for a bit. my unit has stablized at 59.4 for "0" and 57.9 for zone 1. my room temp is about 24c.
I did noticed one spot (~59c) on the back side of the PCB is hot. i also added some pad on the back side to help cool it down.
While doing this... i have managed to lose one screw... lol. do anyone know what size screw i need to replace it with??
1 Like
Those temperatures are well within limits.
I placed the 2 units in the same location. One has the thermal pads and the other doesn't. Both machines are idle and not connected to anything. I left them on for 10hrs.
The one with pads sits at 42C while the other is at 53C.
Logically the device with pads has a hotter case (that's literally the point of it).
I just did this experiment for curiosity. As long as the device doesn't exceed 80C (with or without thermal pads), everything should be fine.
