Best NAS hard drives for Unraid

How Unraid actually stores your data

Unraid doesn't stripe. Every data drive is its own self-contained filesystem, and one dedicated parity drive holds the math to rebuild any single failed disk. Pull a data drive out, read it on another machine, and your files are just there - no array required. That is why you can mix sizes and brands freely, add one disk at a time, and survive a drive dying without losing the pool. It also decides which drive traits matter here and which ones you can stop worrying about.

Parity is the drive that has to be good

Parity does the work nobody sees. Every write to any data drive also updates parity, and when a disk dies, Unraid rebuilds it by reading every remaining drive and writing the replacement from the first sector to the last. A monthly parity check does the same full-disk sweep. So parity takes the most sustained, whole-disk writes in the system, and it has to be at least as large as your biggest data drive, because a 16TB parity can't cover a 20TB disk. Most people buy their largest drive as parity first and grow the data drives underneath it.

Where SMR bites

SMR drives pack more onto a platter by overlapping tracks like roof shingles. The cost shows up on rewrites: to change data in a shingled zone, the drive has to rewrite the whole zone. To hide that, an SMR drive keeps a small conventional buffer and shuffles data into the shingled area later, in the background. Short bursts feel normal. A long, sustained write does not. Once that buffer fills, write speed falls off a cliff and stays down while the drive digs itself out.

Now picture that drive in a parity slot, or as the disk Unraid is rebuilding. Both are hours of continuous, whole-disk writing, which is the one workload SMR handles worst. Throughput craters, latency climbs, and if the drive stalls long enough Unraid or the disk controller decides it has failed and drops it from the array. Lose a drive that way in the middle of a rebuild and you can lose the rebuild with it. None of this is theoretical - it is the same trap that sprung in 2020, when WD quietly slipped SMR into Red NAS drives that then couldn't finish a resilver.

So can you ever run SMR in Unraid?

For parity, no. Not worth discussing. For a data drive that mostly holds cold files you write once and rarely touch, an SMR disk will run fine day to day, because normal writes are light and hit one drive at a time. The bill comes due later: the day that disk fails and has to be rebuilt onto, or the day you copy a few terabytes onto it in one go. The price gap between SMR and CMR is small enough now that most Unraid builders skip the gamble and run CMR throughout, and that is the recommendation here.

Compatibility: there is no list

Unraid doesn't publish an approved-drive list and doesn't care what brand or model you run. No "unverified drive" warning, no vendor lock, nothing to cross-check - if a drive is CMR and healthy, it works. That is why shucking is so popular here: pull a cheap drive out of an external USB enclosure, drop it in a bay, and Unraid treats it like any other disk. The only things you're really vetting are recording tech and reliability, which is exactly what the table is for.

What actually matters once you're on CMR

With SMR off the table, Unraid changes what you optimize for. Nothing forces your drives to match, so you're free to chase value instead of uniformity.

Capacity, parity first. Your parity drive sets the ceiling for every data drive you'll ever add, so buy the biggest one you can justify today and let the array fill in beneath it.

Price per terabyte. This is where Unraid builders actually live. Shucking - pulling a drive out of a cheap external USB enclosure - is popular here for the same reason, and the disks inside are usually CMR NAS or white-label enterprise units. Confirm the exact model before you buy, since a shuck is a gamble on what's inside.

Real failure rates, not rated ones. MTBF is a spec-sheet promise. Backblaze publishes what actually happens, from a fleet of hundreds of thousands of drives, and that annualized failure rate sits on every drive in the table below, against a fleet average near 1.34%. Plenty of consumer models have no Backblaze data at all, which is normal rather than a warning.

Noise and power, if the array lives in earshot. A rack of enterprise drives is louder and thirstier than NAS-class ones. Idle and seek noise are listed below wherever the maker publishes them.

The short version

Run CMR everywhere and treat parity as the one slot you never cheap out on: the biggest drive, the best failure rate, and never SMR. For data drives, buy on price per terabyte from that same CMR pool and mix capacities as you grow. The list below is sorted by measured failure rate, so you're starting from the drives least likely to make you lean on that parity disk in the first place.