I have been using an old PC as NAS and running docker containers (Immich, Nextcloud, paperless-ngx etc.). I got 5 3.5" HDD disks and a Nvme. Even on idle, I am consuming 50-60w. A friend of mine is selling a Qnap NAS which is a dedicated machine and probably consumes less power, although I don’t know if it’s worth it.
In my experience using a PC as a NAS, the power draw isn’t necessarily the drives as they spin down when idle.
I have an old desktop setup as NAS - with 2 drives or eight drives, idle power draw is virtually the same, about 100w, regardless of the OS (Windows, Linux, UnRAID, Proxmox).
I also have an old consumer NAS, with five 4TB drives, and it idles under 20w (I think last I checked it was ~15w… I need to check it again and write that down).
Two very similar systems, one designed to be a NAS, the other a desktop. It really comes down to the motherboard design and capabilities.
I also have a Dell SFF that idles at about 15w, regardless of drive count - one drive or four (and to get four I added a SATA expansion card and rigged some power splitters, really pushing the power supply). That box idling the same, even when pushed well past design, is pretty telling.
And don’t think that SSD drives would do better - spinning disk drives generally have far better idle power than SSD does, and usually much better write power consumption.
So it really depends, and mostly on the motherboard itself. Yes, you’ll get more power usage with more drives, but that’s at write and read time. My SFF idles at 12w, peaks at 80w when converting videos, the read/write power is negligible, same with the NAS (I transfer hundreds of gigs between them every few days).
Good point. I looked mostly at the spec sheet from the manufacturer and for example the Samsung 870 Evo vs Seagate IronWolf NAS Drive. Side note, AFAIK NVME drives have a higher power consumption. Especially PCIe 5.0.
My NAS with 2 HDDs from Seagate has a total powerdraw of around 30-40w. And I don’t spin the drives down.
I wonder if they can be “spinned down” like hard drives. their startup time would be much faster, so it’s shutdown could even be on a tighter schedule. I mean probably they dont have an internal idle timer, but who cares if you can just have something like hd-idle that shuts it down according to a better schedule.
that’s not always the default setup, especially with enterprise drives. also if you have some kind of monitoring, that can keep the drives from going down (for that, use linux hd-idle instead of drive internal idle timer), and it can also wake them up (for that, prometheus node exporter’s smart collector first checks whether a drive is up, and only then collect stats). Interestingly, checking temps with smartctl always spins up my drives, while linux hwmon can give me live temp stats even while the drives are down