The cheapest SSD that won't die in a NAS

Most NAS builds do not die of a bad drive. They die of a cheap drive in the wrong slot, on a machine that loses power without warning.

We run two of these boxes: a 2011 laptop board with a 1.6 GHz Celeron, two cores and 4 GB of RAM, and a small mini-PC with an i5-2520M. Both are still on inexpensive drives. Everything below is either a number we measured on them or a spec we refuse to buy without — and we say which is which.

The short answer

There is no fifth tier. Above the "cheap floor" you are buying warranty length and a workload number, not speed you will ever feel on a home network.

They do not die from write endurance

Take a 500 GB TLC drive rated at 300 TBW. A busy home NAS with photos, backups and a few containers writes maybe 5 TB a year — that is already a high estimate. 300 divided by 5 is sixty years. Even with 3× write amplification from a copy-on-write filesystem, you are still past twenty.

The drive will be obsolete, and the machine it sits in will be scrap, long before the NAND wears out. TBW is a warranty term, not a lifespan prediction. Buying a higher TBW rating to "make it last" is buying an insurance policy against the one failure mode you were never going to hit.

The 256 GB SATA drive in our smaller box reports 8 power-on hours, zero reallocated sectors, and a wear-leveling count of 0. It is new, and that is the point: almost every consumer SSD you can buy today starts its life with a wear budget far larger than the machine around it. What it does not start with is protection against being unplugged mid-write.

What actually kills them

Power loss on a drive with no power-loss protection. Consumer SATA drives almost never have it; the capacitors that flush the mapping table on power failure are an enterprise feature. On a DRAM-less drive the mapping table lives in a small on-chip buffer, and a hard cut during a write is the one moment where losing it matters. Most drives recover. Some do not, and the ones that do not take the whole array with them.

The SLC cache running out. This is where cheap QLC drives fall apart. An empty drive writes fast because it is writing into a fast staging area. Fill that area and sustained writes drop to tens of megabytes per second — slower than the mechanical drive you replaced. A NAS workload is exactly the pattern that triggers it: copy 40 GB of photos in one go and watch the number collapse halfway through.

Heat. A 2.5" SATA drive in a bay with any airflow is fine. An M.2 stick wedged behind a drive cage with no fan is not, and the drive will quietly throttle rather than tell you. Our SATA system drive reports 45 °C under normal load.

The bridge, not the drive. Cheap USB enclosures and add-in SATA controllers fail more often than the disks inside them. The 3 TB drive in our larger box accumulates interface error counts while completely idle, at roughly 0.1 per second, and the count does not grow with I/O volume — that is a firmware or link behaviour, not a dying platter. The lesson generalises: when a drive misbehaves, suspect the cable, the enclosure and the controller before you blame the NAND.

Firmware. A small number of models have shipped with firmware that drops off the bus under specific kernels. Burn a new drive in for 24 hours with real data on it before you trust it with the only copy.

What we measured on our own two boxes

Read that table again. The SSD moves data 45 times faster than the network port it feeds. Swapping in a drive with 10% more sequential bandwidth would have changed nothing you could perceive.

The 100× improvement came from queue depth, not throughput. The old 5400 rpm drive choked on a few dozen concurrent small reads — the JavaScript, the CSS, the fonts — and each one waited behind the last. An SSD serves them in parallel. That is the entire mechanism, and it is why the cheapest drive with a DRAM cache already delivers it.

Practical consequence: on any home NAS, once you are on an SSD, paying more for a faster SSD buys you nothing measurable. Pay for the two things that do not show up in a benchmark — a DRAM cache and a printed TBW — and stop there.

The five lines on the spec sheet

1. DRAM cache. The datasheet says "DRAM" or names a cache size. If it says nothing, assume there is none. DRAM-less drives lean harder on the host and on their own buffers, and they degrade faster as they fill.
2. TLC, not QLC. Capacity per dollar is the QLC selling point, and it is the wrong trade for a machine that receives bulk writes. A useful shortcut: for the same capacity, a QLC drive's TBW is usually around half of a TLC drive's.
3. A TBW number that is actually printed. A manufacturer that omits it is telling you something.
4. Power-loss protection — or an admission that there is none. Consumer drives do not have it. This line decides whether you need a UPS, not whether you need a more expensive drive.
5. SMART attributes that are exposed. A decent drive reports reallocated sectors, pending sectors, uncorrectable errors, UDMA CRC errors and total bytes written. Cheap USB bridges hide all of them, which means you cannot see a drive dying until it is dead.

The configuration we actually run

• System disk: a small TLC SATA SSD with a DRAM cache, 240–512 GB. It holds the OS, the container images and the databases. Nothing else.
• Data disk: a mechanical drive, CMR and not SMR, formatted btrfs or ext4. Photos, media, documents, backups.
• SSD never holds cold data. A drive that is only read from does not need to be fast, and a platter gives you far more capacity per unit of money and per unit of risk.
• Check for SMR before you buy a big 2.5" drive. Many high-capacity portable drives are shingled, and they fall out of an array during a rebuild — exactly when you cannot afford it.

This split is unglamorous and it is what both our machines do. The SSD absorbs everything random and small; the platter absorbs everything sequential and large.

Three commands to run before you trust a drive

# 1. sequential read, bypassing the page cache
sudo hdparm -t --direct /dev/sdX

# 2. the SMART attributes that matter
sudo smartctl -A -d sat /dev/sdX

# 3. 4K random read - the closest thing to a real NAS workload
sudo fio --name=r4k --rw=randread --bs=4k --iodepth=32 --size=1G --runtime=30 --filename=/dev/sdX

We have run the first two on our own drives. We have not run the third on the drive that is in production, so we are not going to invent a number for it — the command is the honest part of this section.

What to look at when the SMART table comes back:

• 5 Reallocated_Sector_Ct — anything above zero, stop putting new data on it.
• 187 Reported_Uncorrect and 197 Current_Pending_Sector — same rule.
• 198 Offline_Uncorrectable — same rule.
• 199 UDMA_CRC_Error_Count — this one climbing means the cable, the backplane or the bridge is at fault, not the NAND. Replace the cheap part first.
• 241 Total_LBAs_Written — a sanity check on how much life has actually been used.

The cheapest reliability upgrade is not an SSD

Our larger box died three times in ten days. The drive was not the cause. The kernel is configured not to reboot on a panic, and the hardware watchdog is disabled, so when the machine hangs it hangs forever instead of recovering.

The third time, the smart plug it is connected to reported 40 W. Idle is 20 W. Power was still flowing — the machine had not lost power, it had stopped responding, and we cut the plug to restart it.

Every one of those events was an unclean shutdown in the middle of writes to an SSD that has no power-loss protection and never will.

So the arithmetic is simple. A cheap drive plus a UPS is more reliable than an expensive drive without one. If your budget only covers one of them, buy the UPS. If you cannot buy a UPS, at least do this: enable "Restore on AC Power Loss" in the BIOS, and put the box on a plug you can switch remotely — the same plug that told us 40 W is the one that let us restart it from another room.

A laptop used as a NAS has this built in. Ours has a 43 Wh battery that turns a power cut into a non-event.

What we would not buy

• No-name drives with no TBW and no SMART. You cannot monitor what you cannot see.
• QLC marketed on cache speed. The benchmark is real; the workload that follows it is not the one they tested.
• Used enterprise drives with 50,000 power-on hours. The remaining life is unknowable and the seller knows it.
• A USB enclosure as permanent storage. Controllers drop off the bus, and a btrfs array that loses a device mid-write is a bad afternoon.
• An M.2 drive with no airflow in a case that was designed for platters.

How this guide was made

Every number above came from our own hardware, captured with hdparm, dd, smartctl, fio where noted, and the host's own counters. No manufacturer supplied a sample, no listing on this page is paid for, and we do not publish prices — they move weekly and a stale one is worse than none.

Last verified 2026-10-18. When we re-measure, or when a drive in one of our boxes fails and we learn something, this date changes.