My guess is, that those 5-6W on idle that gets thrown around in so many blog posts is probably just the CPU in idle state and not the whole board with io and other stuff.
Well, no.
My Futro has some old thin client AMD CPU and the HP Elitedesk g3 mini has an i5 6500T. Both idle at 5-7W (the HP is slightly higher).
My optiplex is just an SFF, so still a regular CPU (i5 6500, without the T), regular SFF power supply, etc. And that one draws 10-15W.
I use a cheap watt meter and the values from above are from that meter.
10-15W is what my Dell optiplex idles at.
I guess, it also depends on your load. My measurements were made at “basic setup idle”, so smb, k3s with a handful of idling containers, pihole. Since I’m the only user of these services and don’t use them that much, the load average is often enough way below 1 (at 4 cores). It’s absolutely possible that someone with higher demands and higher loads pulls more power on average.
So, basically don’t use it.
I mean, I get the effort, but if you account for your own labor, the power savings probably will probably take years to amortise - even with high energy costs like here in Germany.
Just as a rule of thumb: 1W of constant power costs is at most 25cents (40cent/kwh, which is even high for Germany).
My pi 3 idles at 3-4W, my Futro at around 5-6W. I’d say that’s pretty close.
Second that.
RPis are still unbeaten in terms of power usage, but the difference to a low end thin client is small.
Cost isn’t really an argument either, you can get thin clients with case, psu and SSD for something like 40€.
Ebay. Or whatever is your preferred local classified app.
You could try to redo the copy and monitor the system in htop, for example. Maybe there’s a memory or CPU bottleneck. Maybe one of your drives is failing, maybe you’ve got a directory with tons of very small files, which causes a lot of overhead.
3mb/s sounds more like there is something else going on.
Here too there are misconceptions!
What’s important are the hard numbers, soft metrics like user count are misleading! Some may look large at first, but hardly grow with higher engagement, while in others engagement greatly increases the size.
Thing is, you have to measure from the user base on the underside, this graphic obviously uses the wrong method.
I think, I have to save a default-comment for this. Old thin clients are incredibly cheap, small and full blown x86 machines.
The update process is absolutely horrible, especially with containers.
I seriously cannot understand how this hasn’t been fixed ages ago. Upgrading is kind of important and nextcloud isn’t doing that much weird stuff that it didn’t upgrade itself.
Yeah, I’m also not entirely what to think about the machine. There are definitely some quirks that are 100% Dell doing its own thing instead of using standard parts. On the other hand, the case is extremely easy to disassemble, you hardly need any tools.
Yes.
That sounds just stupid enough to actually work.
But then I would need an ssh-enabled wackity whack in my rack.
Tried that. It found one fan, and that’s the GPU fan - which spins down just fine, so no gain here.
…or I could just find a way to fix the fan.
On my machine it has one screen: no battery found…
Linux is not an option. I tried, see my other comment.
How exactly are you “losing” files?