Vector NVMEs keep overheating

post by lminer on Jan 9, 2024

I have a few 4-GPU vectors and have sporadic problems with the SSDs overheating and crashing the system. Has anyone else experienced this, and if so, do you have any fixes? A preferred aftermarket heat shield or an extra fan for example?

post by cody_b on Jan 11, 2024

How did you trace the crashing to overheating SSDs?

post by lminer on Jan 11, 2024

Through a combination of looking at logs and monitoring via munin.

post by Nicholas_C on Jan 11, 2024

I have seen from reviews that the Sabrent heatsink has decent improvements for performance/cooling, but it is atypical that you’re experiencing crashes from the NVME drive, especially on multiple Vectors. It would be quite helpful both to us at Lambda and to other users if you could provide some of the logs/data that points you towards the drives on multiple machines experiencing overheating issues. Also, are these machines plugged directly into the wall or are they perhaps plugged into a power strip/surge protector/UPS device? We commonly see crashes due to these peripheral devices.

post by lminer on Jan 12, 2024

The machines are plugged directly into the wall. It’s a little difficult to provide the datapoints as the logs are now gone. journalctl showed an ssd overheat warning just before the crashes occurred. Do you know which Sabrent heatsink in particular is good for cooling in a vector?

post by Nicholas_C on Jan 12, 2024

Unfortunately I don’t have the review on hand so I don’t remember what model heatsink it was. I’ve also not tried it or had a customer try it in a vector, so I can’t speak to how easy it is to install in the Vector. If you try it and it works/doesn’t work, please update us on the result as it will be helpful for future customers if they run into a similar error.

post by lminer on Jan 12, 2024

I think I found the review. Do you have a sense of what height I can sustain for the heatsink under the GPUs?

post by Nicholas_C on Jan 12, 2024

Oh if the NVME in question is under the GPUs then the Sabrent will not fit unfortunately. For under GPU NVMEs you’ll need a flat heatsink.

post by lminer on Jan 12, 2024

Are there other slots on the motherboard that aren’t underneath the GPU? I could try moving it.

post by Nicholas_C on Jan 12, 2024

What motherboard is it? WRX80 or TRX40?

post by lminer on Jan 12, 2024

I believe it’s a TRX40