I Bought Two TerraMaster NAS Units Before Building My Own Unraid Server. Here's What I Learned.

Most people who end up building a home server didn't start there. They started where I did — looking at Amazon, seeing a TerraMaster for $250 and thinking: that's all I need. Two drives, a NAS OS, done. Maybe I'll run Plex on it.

I bought two TerraMasters. Neither was the answer. What finally was the answer cost more upfront, runs Unraid 7.2.2, and has an i5-13600K in it — which is a genuinely absurd processor to put in a NAS box, and also exactly right for what I'm actually doing with it.

Here's the full story, including what fell short, what made me decide to build, and every part I chose and why.

Attempt One: TerraMaster F4-212

The TerraMaster F4-212 looked right on paper. Four bays, which I figured would give me room to grow. Quad-core CPU. It ran TOS — TerraMaster's own operating system — which has a Docker-lite environment baked in.

Here's what I didn't pay enough attention to: 2GB of RAM. That's the spec that should've been a flashing warning sign. The moment you start running more than one or two Docker containers alongside a media server, 2GB becomes the ceiling you're constantly bumping against.

File storage? Fine. Totally fine. If you want a box that holds drives and shows up on your network, the F4-212 does that without drama. But I didn't just want a drive enclosure. I wanted to run Jellyfin, maybe a VPN container, maybe something else. The minute you start stacking services, 2GB RAM is a problem you can't solve without buying a different unit.

The other thing I underestimated was how much I'd care about the software. TOS works. It's not bad. But it's also clearly a system designed to keep you inside a walled garden. The app ecosystem is limited, the community is smaller, and when something breaks or you want to do something slightly outside the expected path, you're mostly on your own.

I sold it and tried again.

Attempt Two: TerraMaster F2-423

The TerraMaster F2-423 was a step up in CPU — N5095 quad-core, 4GB DDR4, dual 2.5GbE ports. Genuinely better hardware. The processor is no joke for a NAS-class device, and 4GB RAM actually gave me some breathing room.

But I'd made a different mistake this time: I went down to two bays. I told myself I was simplifying. What I was actually doing was buying around the problem I'd already identified and finding a new ceiling to hit.

Two bays on a NAS is not enough if you take storage even moderately seriously. With two drives you're either running RAID 1 (mirroring, so you lose half your capacity for redundancy) or JBOD (no redundancy, living dangerously). I filled those two bays with 10TB drives and immediately started thinking about expansion — which I couldn't do without replacing the unit entirely.

The F2-423 is currently unavailable on Amazon, which may or may not be relevant — but it was a solid unit for basic use. If all you need is a two-bay box for backups and you'll never touch Docker, it's worth considering when it's back in stock. For what I was trying to build, it wasn't the answer.

The Decision to Build

I started actually writing down what I wanted the machine to do:

• Run Jellyfin with hardware transcoding — multiple streams simultaneously
• Run several Docker containers alongside it without fighting for resources
• Have room to add drives as my storage needs grew
• Run a VM occasionally, without the whole system crawling
• Be something I could actually troubleshoot, upgrade, and understand

When I looked at that list honestly, no pre-built NAS unit at a reasonable price point checked all of those boxes. The ones that came close (Synology DS923+, QNAP units with more RAM) were priced close enough to a custom build that the math stopped working in their favor.

The other thing that pushed me toward building was Unraid specifically. I'd been reading about it for a while. The core idea — a parity-protected array where you can mix drive sizes and add storage one drive at a time without rebuilding the whole array — is genuinely different from how consumer NAS software works, and it's the right architecture for how people actually expand storage over time. The Docker and VM integration is also mature in a way that TOS never quite got to.

The Build

Here's every part I chose, with the honest reason why.

On the CPU choice

The i5-13600K is a lot of processor for a NAS. 14 cores, up to 5.1GHz — most of the time it's sitting at low single-digit CPU percentages doing essentially nothing. That's fine. The reason I picked it is what happens when I'm transcoding 4K content in Jellyfin while two other containers are doing background work: nothing slows down. Zero throttling, zero frame drops. The headroom is the feature.

Quick Intel AV1 hardware transcoding on the 13th-gen iGPU is also genuinely fast. If you're running Jellyfin and your clients need transcoding rather than direct play, the Quick Sync implementation on this chip handles it without breaking a sweat.

On the case

The Fractal Design Node 804 is the right answer for this build and I don't think it gets enough credit in home server discussions. It's a cube-style mATX case designed around airflow and drive density. Eight drive bays — more than I need right now, which is the point. The cable routing is sensible, the included fans are quiet, and it sits on a shelf without looking like a mistake.

On the storage

I'm running one parity drive and one data drive — both Seagate Enterprise Capacity 10TB, renewed. That gives me roughly 9.1TB of usable storage. Right now I'm sitting at 87% capacity, which means expansion is coming. The thing I like about Unraid's architecture is that I can add a single drive to the array without touching anything else. No rebuilding, no migrating — just assign the new disk and it's part of the pool.

The renewed enterprise drives have been running without complaint. SMART health is passing on both. Enterprise drives are designed for always-on workloads, which is exactly what a home server is.

Running Unraid 7.2.2

I'm on Unraid 7.2.2 and have been since it released. The Docker integration is the thing that makes this a fundamentally different experience than running TOS on the TerraMasters. Every container gets its own network stack, the Community Applications plugin makes finding and installing containers simple, and when something goes wrong you have actual logs and actual tools to figure out why.

The web interface has improved a lot over the past few versions. It's not pretty in a modern UI sense, but it's functional, everything is where you expect it, and it's fast.

Uptime at time of writing: just over a day — I'd recently rebooted to apply a kernel update. Before that it had been running continuously for several weeks. It mostly just sits there and works.

What I'd Do Differently

Not much, honestly. The build is performing exactly the way I wanted it to. If I were starting over:

• I'd skip the TerraMasters entirely and ask myself the "what does this actually need to do" question first.
• I might look at a slightly smaller case — the Node 804 is perfect for this build but it's a big footprint on a shelf.
• I'd start with more than one data disk. One parity and one data means you're using 50% of your raw capacity for protection. Two data disks is the more sensible starting point.

Should You Buy a Pre-Built NAS or Build Your Own?

The pre-built NAS units aren't bad products. They're the right product for a specific use case — one that wasn't mine. If you're reading this, there's a decent chance it's not yours either, or you wouldn't be looking into it this carefully.

The custom build costs more upfront. You also end up with a machine that does exactly what you need it to do, can be upgraded one component at a time, and doesn't have artificial software limits on what's allowed to run on it.

For me, that math is obvious. Your situation may be different.

— Josiah
Running Unraid since before it was stable. Still running it.