I am using NVMe Base Duo together with two Lexar NM710 1TB configured in RAID1.
It works fine but when I started copying large amounts of data (100+GB) they started heating up like crazy, staying above 70C, with highest recorded temp being 87C. Is this normal? How can I bring temps to a more resonable level? The entire board was too hot to even touch for too long but everything worked as intended.
Idle temps are fine, sitting in the 30-40C range
You should think about cooling. There are heatsinks for M.2 NVMe-SSDs available, but they only help if there is air-flow (passive or active).
BTW: I think that about 70°C is the upper limit of the normal range (depends on your make), but 87°C seems a bit high. So if you regularly copy large amounts of data, you definitely should do something.
That’s what I was thinking about but pimoroni does not make it easy, sandwiching the ssds between their board and the pi. I have actually tried holdung some random 3V fan I had against the side of the pi and this immediately cooled the ssds down to normal temps. But how do you even mount a fan to the side of the pi, there aren’t any easy mounting options as far as I’m aware.
Mounting should be possible. At least if you can design an print using a 3D printer. Something like an extended base-plate with room for the Pi5+Duo and the fan on the side.
Another option is to put the Pi5 and NVMe-duo side-by-side.
@alphanumeric: could you re-post your image of your setup? Something like that should also work here.
I don’t see any extra cooling in the pics, do you just cool passively?
Yes I just cool passively. It’s not a DUO, it’s just the Pimoroni NVMe Base.
One other question: why are you using RAID1? This is for high availability and useful if one of your disks fails. For RAID1 to make sense, you would need a good monitoring that tells you about the problem and then you would have to be able to replace the faulty disk while the system keeps running - something which is mechanically difficult with the Pi5. And you would have to know how to stop a raid, decommission one disk and rebuild the raid after swapping the faulty disk. I think having a good backup and restoring it after a disk dies is much easier.
This is only indirectly related to your thermal problem, but remember that the system has to write every byte twice, so you double the writes and this of course increases the thermal stress.
I’m running a self-hosted cloud based on Openmediavault (that is also managing my RAID) and want max reliability. I’m also doing weekly backups to Google Drive with Rclone.
With RAID1 I assumed that since it’s just data duplicating, you don’t have to rebuild anything, because each drive contains an identical copy of all data.
And yes, I’m aware that I’m hitting both drives at the same time and getting double the heat, it is what is it tbh.
Currently modeling a bracket to hold a fan up to the pi, tried load testing with the fan on and the temps are in the 50-60C ballpark, which is quite acceptable.
This is only partly true. You do have an identical copy, but once one disk fails and you replace it, you have to rebuild the raid. But since you are running Openmediavault, that should be hopefully automatic.
That sounds good.
BTW: I am using a small 10" rack, which already has a mount for a fan:
This image shows an older version, but the rack is now also available with additional space for storage attached below the Pi5.
Ended up modeling a side bracket for a 20mm fan.
Model: pi-bracket - Google Drive
Nice, thanks for sharing!
