LUNAROPS · OPERATIONAL UPLINK 100% UPTIME 1,247d POSTS 893 JEFF.MOON@LUNAROPS.DEV UTC --:--:--

Yxk Zero1 4-Bay NAS: OS Options, Use Cases, and Drive Recommendations

nashomelabtruenasintel-n100self-hostingstorage

The homelab NAS market has quietly gotten interesting at the low end. Devices like the Yxk Zero1 land somewhere between a consumer plug-and-play box (think Synology DS423+) and a full DIY build, offering a real x86 platform with an Intel N100, dual 2.5 Gigabit Ethernet, and 8 GB of RAM in a diskless desktop chassis — all for roughly $170–$200 before drives. That combination unlocks options that ARM-based consumer NAS units simply cannot match: you can install TrueNAS SCALE, Proxmox VE, Unraid, OpenMediaVault, or plain Debian, and the OS will behave exactly as documented because you are running on generic x86_64 hardware with a real UEFI BIOS. The Zero1 is also marketed under the Iron Cow brand (Iron Cow NAS-ZERO 1 / Zero1 Pro), and community members on the TrueNAS forums have confirmed that TrueNAS SCALE 25.10 installs and runs without unusual friction.

Out of the box the Zero1 boots its own bundled vendor OS, so if all you want is a five-minute setup and a phone app for photo backup, you already have one — and the first OS option below covers living with it. But the reason to buy an open x86 box like this instead of a sealed consumer unit is the freedom to run something else. It is a NAS for someone who wants to decide what their NAS runs, has opinions about ZFS, and wants hardware-accelerated transcoding without paying for a Synology Plus series. For that person, the Zero1 Pro is a compelling starting point — provided you go in with eyes open about the RAM ceiling, drive selection, and the real costs of each OS.


The Hardware

The heart of the Zero1 is Intel’s Processor N100, released in Q1 2023 as part of the Alder Lake-N family. The chip has four efficiency cores (no performance cores), runs at up to 3.4 GHz under turbo, and carries a 6 W TDP — the same thermal envelope as an aggressively power-managed laptop. For a device that will run continuously in a closet, that matters enormously. It supports DDR4-3200, DDR5-4800, and LPDDR5-4800 in a single-channel configuration with a 16 GB maximum, and it includes AES-NI hardware acceleration for encryption-heavy workloads.

The iGPU is Intel UHD Graphics with 24 Execution Units (Gen12 / Xe LP architecture), the same graphics block found in other Alder Lake-N parts. It supports Quick Sync Video, which means hardware-accelerated encode and decode of H.264, H.265/HEVC, AV1 (decode only), and VP9 through the i915 kernel driver. On Linux, this surfaces as /dev/dri/renderD128 and /dev/dri/card0. Practical consequence: a single N100 can sustain multiple simultaneous 4K HEVC transcodes in Jellyfin or Plex with CPU utilization that stays well below 50%, whereas software-only transcoding of a single 4K stream saturates the chip. See the Jellyfin vs Plex vs Emby comparison for a deeper comparison of those media servers.

The networking is the other headline feature: dual 2.5 Gigabit Ethernet ports. On a gigabit home network this is mostly headroom, but if you have a 2.5 GbE-capable switch (prices on unmanaged 2.5 GbE switches have dropped substantially), you get full-duplex 2.5 Gbps to a single client — enough to saturate most spinning-rust sequential reads. With LACP 802.3ad bonding or active-backup failover configured at the OS level, you can present both ports as a single logical interface.

The 4K HDMI output is an underappreciated feature. Most competing NAS boxes have no display output at all. Here you can plug in a monitor and keyboard during initial setup without needing a separate machine for SSH access, which simplifies troubleshooting considerably. Under Proxmox or bare Debian, you could also use this as a lightweight desktop workstation — the N100 is fast enough for light productivity tasks.

RAM and upgradeability. The base Zero1 ships with 8 GB. Community reports and the TrueNAS forum review indicate the RAM is in a single SO-DIMM slot (DDR4 non-ECC on the standard model; the Pro variant has been marketed with DDR5), meaning it can be replaced but not supplemented. The maximum supported by the N100 is 16 GB in a single channel, so the upgrade path is one SO-DIMM swap. The M.2 situation differs between variants: the Zero1 Pro explicitly lists two M.2 NVMe slots, making it straightforward to dedicate one to the OS and one to a ZFS L2ARC or Unraid cache drive. The base Zero1 may have only one. Verify before ordering.

Power draw. Based on measurements from comparable N100 NAS platforms: expect roughly 10–15 W for the board and processor at idle with no drives spinning, and 6–8 W per spinning hard drive added on top. A fully populated 4-bay system with four NAS-rated HDDs in active use lands in the 38–45 W range under light NAS load, dropping to around 28–35 W with aggressive HDD spindown timers. An all-SSD build would idle closer to 15–18 W total. For a device left on 24/7, even the difference between 30 W and 45 W is about $13/year at $0.12/kWh, so spindown timers are worth configuring but not agonizing over.


OS Options

Choosing an OS for the Zero1 is consequential in a way it is not for a Synology — you are choosing a fundamentally different data model, upgrade path, and support ecosystem. Here is an honest assessment of each option for this specific hardware, starting with the one most reviews skip: keeping what it shipped with.

Stock OS (the bundled Iron Cow / Yxk software)

The Zero1 arrives with its own preinstalled, vendor-maintained NAS OS on internal boot media — marketed under Iron Cow’s “AI-NAS” branding — and out of the box it behaves like a budget Synology clone. You get a browser-based management portal, companion apps for Windows, macOS, mobile, and TV, remote access that works without touching your router, and a feature set built around consumer use cases: an AI-tagging photo album, photo and file management, a small app catalog, and SMB file sharing. For a meaningful fraction of buyers this is genuinely all they need, and it is the only option here that asks for zero Linux knowledge, zero RAM upgrade, and zero license fee.

One honest caveat up front: the exact name and lineage of this bundled OS are not clearly documented in public sources. These budget x86 NAS boxes typically ship a proprietary or lightly-rebranded vendor build, and the identity can shift between production batches. Check what is actually on your unit, and download the recovery image from the vendor’s software/update page before you change anything — if you wipe it to try TrueNAS and want to go back, you will need that image.

The case for keeping it is that it works today. The mobile photo-backup-and-AI-album experience is the one feature people most often miss after moving to TrueNAS or Unraid, and the stock OS delivers it for free with a polished phone app and vendor-relayed remote access — no reverse proxy, no Tailscale, no certificate wrangling. If you bought the Zero1 to replace a pile of USB drives and back up a camera roll, this is the path of least resistance, and you can always reflash later.

The downsides are the flip side of that convenience:

  • Closed and unauditable. It is proprietary; you cannot inspect what it does, and budget NAS firmware has a poor track record on default credentials, phone-home behavior, and unpatched CVEs. Treat the device as untrusted: isolate it on its own VLAN, never port-forward it to the open internet, and lean on the vendor cloud as little as possible.
  • Vendor-dependent longevity. Updates, security patches, and the remote-access cloud all depend on a small company continuing to ship. If they stop, there is no community to pick it up — unlike TrueNAS, OMV, or Debian, which outlive any single vendor.
  • No real data integrity. These OSes typically layer ext4 or btrfs over mdadm or a simple proprietary volume manager. There is no ZFS end-to-end checksumming, so silent bit-rot is caught late or not at all — exactly the failure mode the ZFS case in the ZFS homelab storage post exists to prevent.
  • Lock-in and migration cost. The on-disk format is proprietary enough that leaving usually means copying all your data off to somewhere else first, then reformatting. There is no clean in-place conversion to ZFS or an Unraid array.
  • Rough edges. Translated UIs, thin English documentation, and a small third-party app selection next to TrueNAS’s catalog or Unraid’s Community Applications.

How it compares: the stock OS is the “Synology-lite” lane. It trades away the entire reason to buy an open x86 box — the freedom to run ZFS, a hypervisor, or any mainstream Linux NAS distro — in exchange for plug-and-play convenience. That is a coherent choice, but note the irony: if convenience is genuinely all you want, a real Synology or QNAP gives you the same turnkey experience on a far more mature, longer-supported OS. The Zero1’s value proposition is precisely the freedom that keeping the stock OS declines to use. Keep it if you want a working NAS this afternoon and you treat the data on it as non-critical or backed up elsewhere; migrate to one of the options below the moment that data becomes something you cannot afford to lose.

TrueNAS SCALE

TrueNAS SCALE is the Debian-based successor to FreeNAS. It runs ZFS natively, ships with a mature web UI, and supports Docker-based apps through its Apps system (backed by k3s/Helm under the hood). It is the closest thing to a purpose-built NAS OS for x86 hardware, and it has verified community reports of running on the Zero1 Pro.

The strongest argument for TrueNAS SCALE on this hardware is ZFS. You get copy-on-write semantics, per-dataset compression (lz4 is enabled by default and is effectively free on the N100), snapshots, send/receive for replication, and checksumming that actively detects silent data corruption. The i915 driver is present in TrueNAS SCALE’s kernel, so Intel Quick Sync works for Jellyfin or Plex deployed through the Apps catalog.

The caveat is RAM. ZFS ARC (Adaptive Replacement Cache) is a read cache that lives in RAM and is designed to consume whatever memory the OS is not otherwise using. On an 8 GB system, ZFS will claim 4–6 GB for ARC, leaving 2–4 GB for the OS, apps, and any running containers. For a NAS doing light SMB/NFS serving with occasional media playback, that is workable. Add a Nextcloud container, a Jellyfin instance with an active library scan, and a few Samba connections, and you will feel the pressure. Upgrading the SO-DIMM to 16 GB before deploying TrueNAS SCALE is the single most effective hardware investment you can make for this platform — not a luxury. See the ZFS homelab storage deep dive for a detailed treatment of ZFS ARC behavior and the 1 GB/TB rule of thumb.

TrueNAS SCALE also has no license cost, which matters when comparing it against Unraid.

Unraid

Unraid takes a fundamentally different approach. Rather than a traditional RAID or ZFS pool, it uses a parity-protected array where individual drives remain independently formatted with XFS or BTRFS. You can add drives of different sizes without reformatting, and losing the parity drive does not lose data — only protection. Docker is the primary app mechanism, and the container ecosystem in the Unraid community app store is extensive.

For the Zero1, Unraid’s main advantages are flexibility with mixed drive sizes and significantly lower RAM overhead. Unraid does not require a fixed RAM allocation for storage metadata the way ZFS does; 8 GB is genuinely comfortable. Quick Sync transcoding works under Unraid with appropriate driver installation.

The honest downsides: Unraid costs money. Current pricing as of mid-2026 starts at $49 for the Starter tier and $109 for Unleashed, with a $36/year renewal for continued updates. That is not outrageous, but it is a real recurring cost that TrueNAS SCALE, OMV, and Proxmox do not impose. Unraid also uses its cache drive tier to smooth over the parity array’s write performance penalty — for random writes, the parity array is slow until a cache drive is present, so budgeting for an M.2 NVMe cache is effectively mandatory for a good experience.

OpenMediaVault 7

OMV 7 is a Debian-based NAS OS that runs on extremely modest hardware, with a lighter footprint than TrueNAS SCALE. It supports Samba, NFS, AFP (deprecated but present), rsync, and S.M.A.R.T. monitoring out of the box, and the openmediavault-compose plugin exposes a Docker Compose interface for running containers. It does not natively include ZFS support in the base install, though the openmediavault-zfs plugin adds it.

The appeal on the Zero1 is RAM efficiency: OMV’s base system uses well under 1 GB at idle, leaving most of the 8 GB available for whatever you run on top of it. If your plan is “SMB shares and a handful of Docker containers,” OMV delivers that with the least overhead. The UI is clean and approachable, making it one of the better choices for a first NAS.

The trade-off is depth. OMV does not ship with the enterprise-grade ZFS feature set that TrueNAS SCALE integrates (dedicated pool management UI, replication tasks, scrub scheduling), and its plugin ecosystem is maintained by a smaller community. It is an excellent choice for simplicity; it is not the right tool if your priority is ZFS feature completeness. For LUKS-based full-disk encryption on an OMV setup, see the LUKS full-disk encryption post.

Proxmox VE

Proxmox VE reframes the problem entirely. Instead of treating the Zero1 as a NAS, you treat it as a type-1 hypervisor and run TrueNAS SCALE, OMV, or any other NAS OS as a virtual machine — while simultaneously running other VMs and LXC containers on the same hardware.

The canonical Proxmox NAS pattern: install Proxmox on the M.2 boot drive, pass through the SATA controller to a TrueNAS SCALE VM using PCIe passthrough, and run your remaining workloads (Home Assistant, Jellyfin, Nextcloud) in separate LXC containers. The HDMI output becomes genuinely useful here — you get a Proxmox console on a monitor without needing a laptop or KVM switch. Intel QuickSync can be passed through to an LXC container or a VM, though GPU passthrough to a VM requires careful IOMMU grouping (verify that your SATA controller and iGPU are in separable IOMMU groups before committing to this topology).

The real limitation is RAM. Running Proxmox itself, a TrueNAS SCALE VM with ZFS, and a few LXC containers on 8 GB is genuinely cramped — TrueNAS needs at least 8 GB to itself for a comfortable ZFS deployment. On the base Zero1 with 8 GB soldered/socketed, the Proxmox topology only makes sense if you upgrade to 16 GB first. See the Proxmox VE comprehensive guide for the full hypervisor setup walkthrough and the KVM/libvirt without Proxmox post if you want raw KVM without the Proxmox layer.

Bare Debian/Ubuntu

Running stock Debian 12 or Ubuntu 24.04 LTS gives you maximum control and minimal overhead. You install zfsutils-linux for ZFS, samba for SMB shares, and build whatever stack you want with Docker Compose, mergerfs+snapraid, or pure ZFS. There is no opinionated UI getting in your way, and you can pin kernel versions to avoid regressions that have occasionally affected Quick Sync support through distribution updates.

The cost is time. There is no NAS-specific web UI for pool management, no guided SMART alert setup, and no one-click app installs. For an experienced Linux administrator who wants complete transparency into what is running, this is rewarding. For someone who wants a working NAS by the weekend, it is not the right starting point. The mergerfs+snapraid combination is worth knowing about as an alternative to ZFS if RAM is tight and you do not need copy-on-write semantics — see the homelab hardware guide for a broader discussion of storage topology trade-offs.

Comparison Table

OS License Cost ZFS Support Docker/VMs QuickSync RAM Requirement Beginner-Friendly Best Use Case
Stock Iron Cow OS Free (bundled) No Vendor app catalog only Vendor app only 8 GB fine Highest Plug-and-play, phone photo backup, zero setup
TrueNAS SCALE Free Native, excellent Docker (k3s) + VMs Yes (i915) 8 GB min, 16 GB recommended Moderate ZFS-first NAS, media server
Unraid $49–$109 + $36/yr Plugin (limited) Docker-first Yes 8 GB fine High Mixed drives, large media library
OpenMediaVault 7 Free Plugin Docker Compose Yes 4 GB fine High First NAS, light workloads
Proxmox VE Free (subscription optional) Via VM/ZFS Full VMs + LXC Passthrough 16 GB strongly recommended Low Hypervisor running NAS as VM
Bare Debian/Ubuntu Free Manual install Manual Manual 4 GB+ Low Maximum control, custom stacks

Backing up the stock OS before you reflash

If you bought the Zero1 and decided to install TrueNAS, Unraid, or anything else, do one thing before you wipe it: take a full image of the factory boot media. The vendor does not publish a guaranteed-available recovery image, batches ship slightly different builds, and the only thing that reliably gets you back to a working stock system is a byte-for-byte copy of what shipped. It takes fifteen minutes and a spare USB drive, and it is the difference between “I’ll just flash it back” and “the download page no longer has my version.”

Separate the two things you might want to preserve. The boot media holds the OS itself — the firmware, the web portal, the companion-app backend. Your actual data lives on the drives in the bays and is a separate concern; copy that off over the network (SMB/rsync) the normal way. This procedure is only about the small internal boot device, so that the operating system can be restored.

Image it from a live USB, not from the running OS. You cannot cleanly image a mounted, running root filesystem — caches and open files make the copy inconsistent. Boot the box from a live Linux USB instead so the stock OS is dormant while you read it. Clonezilla is purpose-built for this; Ubuntu Desktop “Try Ubuntu” or SystemRescue work equally well if you prefer raw dd. Write the live image to a USB stick with Rufus, balenaEtcher, or dd, plug it into the Zero1, and use the 4K HDMI output and a keyboard for the console — no second machine needed.

Enter the BIOS to boot from USB: tap Del repeatedly at power-on (some units use F2, F7, or Esc), move the USB device to the top of the boot order or use the one-time boot menu, and disable Secure Boot if the live image refuses to start. Save and reboot into the live environment.

Identify the OS device — and do not guess. This is the step that bites people. The boot device is small (an eMMC module or a low-capacity M.2/SATA SSD, typically 32–128 GB); the bays are multi-terabyte. List everything with sizes and models:

1
lsblk -o NAME,SIZE,MODEL,TRAN,MOUNTPOINT

An eMMC boot module shows up as /dev/mmcblk0 (with mmcblk0p1, p2… partitions). A small SSD shows up as /dev/sda or /dev/nvme0n1. The four NAS drives will be the obvious large entries — ignore them. Confirm the small device is the one with the OS partitions on it before you go any further. Read the size twice.

Image it. The safe, recommended path is Clonezilla in device-to-image mode, which understands partition tables and only copies used blocks. Pick device-image, point it at an attached USB or network share as the destination, select the small OS disk as the source, and let it run.

If you prefer raw dd (simplest, captures everything including the partition table and bootloader, but copies the whole device including empty space), mount a spare USB drive at /mnt/usb and run:

1
sudo dd if=/dev/mmcblk0 of=/mnt/usb/ironcow-stock-os.img bs=4M status=progress conv=noerror,sync

Substitute the device you confirmed with lsblk. Then compress and checksum the result so you can verify it later:

1
2
zstd -19 --rm /mnt/usb/ironcow-stock-os.img        # -> ironcow-stock-os.img.zst, removes the raw file
sha256sum /mnt/usb/ironcow-stock-os.img.zst > /mnt/usb/ironcow-stock-os.img.zst.sha256

The of= argument is destructive. dd writes wherever you point of= with zero confirmation — name the wrong device and you overwrite a populated data drive instantly. Re-run lsblk immediately before the command and confirm the size matches the small OS device, not a bay. This is why imaging from the small device (if=) to a file is safe, but restoring to a device (of=) demands the same paranoid double-check.

Restoring later. Boot the same live USB, attach the backup, and write the image back to the same device (or an identical-or-larger replacement — never smaller, the partition table won’t fit):

1
2
zstd -d /mnt/usb/ironcow-stock-os.img.zst -c | sudo dd of=/dev/mmcblk0 bs=4M status=progress
sync

Clonezilla’s device-image restore mode does the same thing with guardrails and is the friendlier option if the raw pipe makes you nervous.

Fallback. Also check the vendor’s software/update page for an official recovery or reflash image and save a copy alongside yours. Treat it as a backup to your backup, not a replacement — vendor pages take down old builds, and a community member’s mirror may not match your batch. Your own image of your own unit is the one you trust.


What You Can Do With It

Media Server with Hardware Transcoding

Jellyfin or Plex with Quick Sync is the single most compelling use case for a NAS with an N100. Install either on TrueNAS SCALE via the Apps catalog, or in a Docker container on any other OS, and configure hardware acceleration with the Intel VAAPI backend. The render node path is /dev/dri/renderD128; the container needs access to this device and membership in the render group (GID 104 on Debian-based systems).

1
2
3
4
5
6
7
8
# Docker run snippet for Jellyfin with QuickSync
docker run -d \
  --name jellyfin \
  --device /dev/dri/renderD128:/dev/dri/renderD128 \
  --group-add render \
  -v /mnt/media:/media \
  -p 8096:8096 \
  jellyfin/jellyfin

With Quick Sync active, the N100 can sustain three or four simultaneous 4K HEVC transcode streams. AV1 decode is hardware-accelerated; AV1 encode falls back to software. Confirm acceleration is working with intel_gpu_top — if the Video and VideoEnhance engines show nonzero utilization during playback, Quick Sync is engaged.

NAS and Backup Target

The fundamental use case. Samba for Windows and Linux clients, NFS for Linux-native mounts, and Netatalk or Samba’s macOS AFP emulation for Time Machine. With 2.5 GbE and four populated bays, you have enough throughput for simultaneous multi-client access without the network becoming the bottleneck. For NFS mount options and performance tuning see the NFS and network storage post.

Configure rsync over SSH for pull-based backups from remote machines:

1
2
# Pull a remote directory to the NAS (run on the NAS)
rsync -avz --delete user@remote-host:/important/data/ /mnt/backup/remote-host/

Self-Hosted Cloud

Nextcloud for file sync and collaboration, or Immich for photo management with on-device machine learning tagging (Immich’s ML container runs adequately on the N100 for a personal photo library). The self-hosted Nextcloud guide covers the Nextcloud deployment in detail, including the Redis and PostgreSQL backing services that make it perform well.

Home Automation Hub

Home Assistant runs happily in a Docker container or an LXC container under Proxmox. Its RAM footprint is modest — under 512 MB for a typical home installation — making it one of the least demanding things you can add to the Zero1 stack. Under Proxmox, give it a dedicated LXC with 1 GB RAM and 8 GB disk.

VPN Gateway and Network Appliance

The dual 2.5 GbE ports make the Zero1 serviceable as a WireGuard endpoint or even a soft router running in a Proxmox VM. AES-NI ensures that encrypted tunnels have negligible CPU overhead. A common pattern: one NIC on the LAN-facing VLAN for NAS duties, the second NIC on a DMZ or WAN-adjacent VLAN for the VPN endpoint — full network separation without a second device.

Developer and Lab Server

The N100 has enough headroom for k3s (lightweight Kubernetes) with a handful of workloads, especially if you keep the NAS OS overhead low with OMV or bare Debian. LXC containers under Proxmox are even lighter than Docker and share the host kernel with near-zero overhead, making the Zero1 a capable lab platform for infrastructure-as-code experiments. See the Proxmox VE setup guide for the initial setup and LXC workflow.


Drive Recommendations

ZFS (TrueNAS SCALE)

ZFS has non-negotiable drive requirements. The most important: no SMR (Shingled Magnetic Recording) drives. Device-managed SMR (DM-SMR) drives — including the standard WD Red (non-Plus) — perform catastrophically under ZFS resilver operations. The sustained random writes during a resilver exhaust the drive’s internal translation buffer, causing write amplification that can stretch a resilver from hours to days and risk timeout-related pool degradation. The WD Red controversy from 2020 established this clearly, and WD’s own documentation now flags the non-Plus Red as unsuitable for ZFS. The rule is simple: only CMR (Conventional Magnetic Recording) drives in a ZFS pool.

Recommended CMR NAS drives for the Zero1:

  • WD Red Plus (SATA, CMR, 3.5") — available 1–8 TB, 5400 RPM, 128–256 MB cache. The canonical choice for home ZFS pools. The “Plus” suffix is critical; plain “Red” without Plus is SMR.
  • Seagate IronWolf (SATA, CMR, 3.5") — available 1–20 TB, 5400–7200 RPM. Generally competitive with WD Red Plus on price per TB, particularly at 8 TB and above.
  • Toshiba N300 (SATA, CMR, 3.5") — 4–16 TB, 7200 RPM, 256 MB cache, rated for up to 180 TB/year workload. A solid third option with a 3-year warranty.

RAM and pool sizing with 8 GB. ZFS ARC will consume 50–75% of available RAM by default, leaving 2–4 GB for everything else. With 4 drives in a RAIDZ1 or two-way mirror topology, a typical home data set (media files, documents, backups) does not stress ARC — most reads will be sequential and ARC hit rates on large media files are modest anyway. Where 8 GB genuinely hurts is with many small files (like a photo library or a git repo collection) and with running multiple apps simultaneously. If you are deploying TrueNAS SCALE with Nextcloud, Jellyfin, and active snapshots, upgrade to 16 GB. The N100’s single-channel memory controller limits you to one SO-DIMM slot, so the upgrade is a replacement, not an addition — pull the 8 GB stick and install a 16 GB DDR4 (or DDR5, depending on your Zero1 variant) SO-DIMM.

A 4-drive RAIDZ1 pool layout on the Zero1:

+--------------------------------------+
|         TrueNAS SCALE: ZFS Pool      |
|                                      |
|  [ RAIDZ1 — 1 parity + 3 data ]     |
|                                      |
|  Bay 0: WD Red Plus 6TB (sda)        |
|  Bay 1: WD Red Plus 6TB (sdb)        |
|  Bay 2: WD Red Plus 6TB (sdc)        |
|  Bay 3: WD Red Plus 6TB (sdd)        |
|                                      |
|  Usable: ~16.4 TB  (3 × 5.46 TiB)   |
|  Parity: 1 drive   (survives 1 fail) |
+--------------------------------------+
|  M.2 NVMe (if present): OS or SLOG   |
+--------------------------------------+

Alternatively, four drives in two-way mirrors (2×mirror pairs) gives you 12 TB usable with faster resilver times and better random I/O — at the cost of half your raw capacity to redundancy.

Unraid

Unraid is substantially more tolerant of drive heterogeneity and SMR. Its parity mechanism writes through a cache drive first (typically an NVMe SSD), then moves data to the array in large sequential chunks during a mover pass. Sequential writes to CMR and SMR drives look largely identical, because both handle large sequential write streams well. The problematic workload for SMR — sustained small random writes — is specifically what the cache tier avoids.

Practical Unraid drive strategy for the Zero1: use whatever NAS-rated drives you have or can find on sale for the parity array (WD Red, Seagate IronWolf, Toshiba N300, even older CMR drives of different sizes), and put a 500 GB–1 TB NVMe in one of the M.2 slots as the cache pool. Most user data will live on the cache long enough for the mover to write it to the array efficiently. Budget for the cache drive as a required component, not an optional upgrade.

All-SSD Build

The Zero1’s bays accept standard 3.5" drives, but 2.5" drives with a 3.5" adapter tray work fine. A four-bay all-SSD build using 2.5" SATA SSDs — say, four 2 TB Crucial MX500 or Samsung 870 EVO drives — gives you an 8 TB diskless NAS with roughly 8–12 W total system power draw at idle. The cost premium over spinning rust is significant at this scale: 2 TB SATA SSDs cost roughly $80–100 each, putting the drive cost at $320–400 for 8 TB usable in RAIDZ1, compared to roughly $100–120 for a comparable spinning-rust build with WD Red Plus 4 TB drives. The power savings ($15–20/year) do not come close to amortizing that difference. The all-SSD build is worth considering if silence is a priority (SSDs are completely silent; the Zero1’s case fan is the remaining noise source), if the NAS will host a workload with heavy random I/O (databases, VM disk images), or if you already own the SSDs.

Drive Reference Table

Model Interface CMR/SMR Capacity Approx. $/TB Recommended For
WD Red Plus SATA 6Gb CMR 1–8 TB $18–22 ZFS RAIDZ1/mirrors, all NAS OSes
Seagate IronWolf SATA 6Gb CMR 1–20 TB $17–21 ZFS, Unraid, OMV; best $/TB at 8 TB+
Toshiba N300 SATA 6Gb CMR 4–16 TB $18–22 ZFS, Unraid; 7200 RPM for higher throughput
WD Red (no Plus) SATA 6Gb SMR 1–6 TB $14–17 Unraid cache array only — NEVER in ZFS
Crucial MX500 2TB SATA 6Gb SSD 2 TB $40–50/drive All-SSD build, Unraid cache, OS/boot
Samsung 870 EVO 2TB SATA 6Gb SSD 2 TB $45–55/drive All-SSD build, high-IOPS workloads
WD Red SN700 500GB NVMe M.2 SSD 500 GB $50–65/drive Unraid cache tier, TrueNAS OS, Proxmox boot

Interesting Findings

RAM is DDR4, not DDR5, on the base model. Despite the Intel N100 supporting DDR5, community reports including the TrueNAS forum review consistently describe the standard Zero1 as using DDR4 SO-DIMM memory, not DDR5. The Zero1 Pro branding appears to vary — some listings reference DDR5, others do not. Verify the exact memory type on your specific unit before ordering an upgrade module, as DDR4 and DDR5 SO-DIMMs are physically incompatible.

M.2 slots: base vs. Pro. The Zero1 Pro explicitly lists two M.2 NVMe slots, which is significant: you can use one for the OS boot drive and one for a ZFS L2ARC or Unraid NVMe cache without consuming any of the four bays. The base Zero1’s M.2 configuration is less clearly documented in product listings — it may have one slot or none. Check the product listing carefully, and if you are buying primarily for a Proxmox or Unraid deployment where an NVMe cache matters, target the Pro variant.

TrueNAS SCALE 25.10 confirmed working. A community member posted a unit review on the TrueNAS forums covering TrueNAS SCALE 25.10 on the Zero1 Pro. The conclusion was that the Intel-based hardware with a real BIOS makes installation less fiddly than ARM-based NAS units. No unusual driver issues were noted, and Quick Sync was accessible through the standard Intel i915 pathway.

Quick Sync kernel driver stability. One OMV forum thread documented a case where a kernel update broke Quick Sync for Jellyfin on an N100 platform — transcoding fell back to software silently. The fix was a package downgrade or waiting for the next kernel point release. This is not a Zero1-specific issue but is relevant to any N100 deployment on a rolling or semi-rolling Linux base. TrueNAS SCALE’s controlled update cadence makes it less susceptible to this than OMV’s Debian-tracking approach.

Proxmox and IOMMU grouping. The N100’s IOMMU groupings matter if you plan to pass through the SATA controller to a TrueNAS VM while keeping the iGPU accessible from another container. Some Alder Lake-N motherboards group SATA and iGPU in the same IOMMU group, making clean separation impossible without using the vfio-pci stub driver for both. Boot Proxmox with intel_iommu=on iommu=pt in the kernel command line and run find /sys/kernel/iommu_groups -type l | sort -V to inspect groupings before committing to a passthrough topology.

Power management and C-states. Community N100 NAS builders have noted that enabling C6/C7 CPU sleep states and PCIe ASPM in the BIOS reduces idle draw by 3–5 W with no observable latency impact on NAS workloads. Check BIOS power management settings — many mini PC-derived motherboards ship with these disabled for maximum compatibility.


Verdict

Keep the stock OS if you want a working NAS this afternoon, value the bundled phone photo-backup and remote-access apps over ZFS and control, and treat the data on it as non-critical or backed up elsewhere. Isolate it on its own VLAN, keep an image of the boot media so you can recover it, and go in knowing you are using the box as a budget Synology rather than the open x86 platform that justifies its price. The moment the data matters, migrate to one of the options below.

Buy TrueNAS SCALE if data integrity is your priority and you are willing to upgrade to 16 GB RAM. ZFS’s checksumming, snapshot replication, and compression are genuinely valuable for a NAS that stores irreplaceable data. Budget: Zero1 Pro ($200) + 16 GB SO-DIMM ($35) + four WD Red Plus 4 TB drives (~$380) = roughly $615 for a 10.9 TB usable RAIDZ1 pool.

Buy Unraid if you have drives of different sizes you want to consolidate, or if you are building a large media library and the Docker app ecosystem is more important to you than ZFS. Accept the license cost as the price of a better mixed-drive experience and a more polished container UI.

Choose OMV if you are new to self-hosted NAS, want a free OS with a clean UI, and your workload is primarily file sharing and a few containers. The lower RAM overhead makes 8 GB more comfortable, and the learning curve is gentler.

Deploy Proxmox if you want the Zero1 to be a general-purpose hypervisor that also runs a NAS OS as a VM. Do not attempt this on 8 GB RAM — upgrade first. This is the highest-ceiling option and the most complex.

Run bare Debian if you are an experienced Linux sysadmin who wants no abstraction between you and the hardware, and you enjoy the process of composing a system from first principles.

For most homelab users reading this, TrueNAS SCALE with 16 GB RAM is the right answer — it earns the hardware fully, and the Zero1 Pro’s confirmed compatibility removes the usual uncertainty that comes with running NAS OS software on generic consumer hardware.


Sources

Comments