So I have finally built my NAS. I used an N100 CPU because I saw it has low power consumption.

Right now I have 2 NVMe SSDs and 2 HDDs. I have installed proxmox on the 2 SSDs as RAID1. I have not partitioned the HDDs yet, they are just plugged in and powered on.

Just booting into proxmox, without any VMs or containers running, I am pulling 45W from the wall. This looks super high to me, and I’m afraid that starting to use the HDDs and running some VMs may double this…

I don’t have much references, but I have an Odroid with an external self-powered HDD, it is using 5W. I have a raspberry pi 4 with an external HDD, the raspberry is pulling 3W and the HDD 3W.

With these data, I was thinking I wouldn’t go over 20W. 45W is enormous and not something I can run 24/7, kind of a fail for a NAS…

Have I done something wrong or is it just how much it’s supposed to pull?

Edit: I have come across powertop. Using the auto tune, I was able to drop to 33-35W. I have unplugged the HDDs and dropped to 22W. I guess I cannot go lower, this may be because of the PSU or the 2 NVMe

  • poVoq@slrpnk.net
    link
    fedilink
    English
    arrow-up
    5
    ·
    9 months ago

    For power consumption, NVMe drives use quite a lot of power, especially PCIe 4.0 ones. About 5W each during use.

    3.5" HDD (especially 7200rpm or more) also consume significantly more than 2.5" 5400rpm HDDs that are optimized for low power (the latter use about 1W during use).

    SATA SSDs fall somewhere in the middle.

    • Kwa@derpzilla.netOP
      link
      fedilink
      English
      arrow-up
      2
      ·
      9 months ago

      Thanks. So If I take 5W a disk, I may have 20W just from the disks.

      There are still 25W remaining. For a low power CPU that looks like a lot 😕

      • poVoq@slrpnk.net
        link
        fedilink
        English
        arrow-up
        3
        ·
        9 months ago

        I never really tested it with 3.5" HDDs, but a google search makes it sound like they might be closer to 10W each.

      • rambos@lemm.ee
        link
        fedilink
        English
        arrow-up
        1
        ·
        9 months ago

        My guess would be 25W from disks, 10W from CPU+RAM+MBO and 10W waste from PSU. Just a guess as I said

    • CyprianSceptre
      link
      fedilink
      English
      arrow-up
      1
      ·
      9 months ago

      How does that work out in terms of energy consumption though?

      If NVMe is at least 10x faster, but consumes 5x more power, it will use less energy to read or write the same amount of data overall.

      • poVoq@slrpnk.net
        link
        fedilink
        English
        arrow-up
        2
        ·
        9 months ago

        To answer that question one would need to dive deeply into idle vs. in active use power consumption. It’s not like a NAS gets turned off when not in use.

      • rambos@lemm.ee
        link
        fedilink
        English
        arrow-up
        1
        ·
        9 months ago

        My server use about the same power no matter if its idle or under full load, so I dont think drive speed makes any difference. My SSDs use much less than my HDD anyway

    • BearOfaTime@lemm.ee
      link
      fedilink
      English
      arrow-up
      1
      arrow-down
      1
      ·
      edit-2
      9 months ago

      I’m surprised to hear NVMe us that much power - I had no idea, and just assumed they used very little.

      It sounds like from a power perspective that 2.5" is better, and SSD is lowest power?

      • MangoPenguin@lemmy.blahaj.zone
        link
        fedilink
        English
        arrow-up
        3
        ·
        9 months ago

        They use practically nothing when idle, but it spikes up dramatically under heavy load.

        Still much less on average than a HDD that uses 5-10W even when idle and spun up.

      • poVoq@slrpnk.net
        link
        fedilink
        English
        arrow-up
        2
        ·
        9 months ago

        In my limited testing 2.5 HDD uses the least power, and SATA SSDs are slightly worse than those, but I guess it also depends on the model.