SerpApi.Org
Old HDDs can uncover gigabytes upon gigabytes of essentially free storage. If they're already stuck in a drawer in your home somewhere, you might as well use them. But should you trust them with your files? Debatable.
One way to keep tabs on how these HDDs are doing is with SMART health checks. These are the values I always verify before moving anything important onto an HDD.
Your HDD has been keeping receipts this whole time
It's not flawless, but it helps
SMART (Self-Monitoring, Analysis, and Reporting Technology) is a diagnostic system built into every modern hard drive. Yes, a 10-year-old HDD still falls under that generous umbrella.
The drive constantly logs data about its own health to a reserved area of the platters, tracking things like operating hours, remapped sectors, and read errors. Think of it as a black box recorder that's been running since the day the drive first spun up. Of course, it also applies to SSDs, and I check this data as a rule even with a brand-new SSD.
Checking it takes a minute or two. I use CrystalDiskInfo for this, because it shows every attribute in a single window, but smartctl does the same on any operating system. Each attribute has a normalized value and a raw value, and for everything on this list, the raw value is the one that matters.
There's a catch, because of course there is. SMART tracks over 70 attributes, and most of them tell you very little. Instead of squinting at all the rows of data, make sure you always check the ones I'll talk about below. Five of these are the ones that Backblaze says it tracks, and it has hundreds of thousands of drives to learn from.
SMART 5 tells you the platter surface is already losing sectors
Reallocated sectors are the drive quietly patching itself
SMART 5 counts reallocated sectors, which are spots on the platter where the drive hit read or write errors and then decided to move the data to a spare area instead.
Every HDD ships with a stash of spare sectors for exactly this purpose, so a small number here isn't an automatic disaster. Both Backblaze and Google's classic large-scale drive studies found this attribute to be one of the strongest predictors of failure, though, so it's earned its spot at the very top of my checklist.
What matters more than the number itself is whether it's moving. A drive showing the same handful of reallocated sectors for months is probably stable, but a climbing count means the platter surface is actively degrading and the drive is burning through its spares.
SMART 9 points to where the drive sits on the bathtub curve
Not related to any actual bathtub
SMART 9 counts power-on hours, and unlike everything else on this list, it's not a warning sign at all. It's just context.
HDD failure follows what's known as the bathtub curve: lots of failures early on from manufacturing defects, a long stretch of relative calm, and then a rise as old age sets in.
A 2026 study from Secure Data Recovery, which analyzed SMART data from 5,106 dead hard drives, found that 49.8% of them failed before hitting one year of runtime, or more precisely, 8,760 power-on hours.
These findings changed how I read this value on an old HDD. An HDD showing 20,000 or 30,000 clean hours has already survived the deadliest stretch of its life, which arguably makes it safer than a brand-new one. The curve does bend back up eventually, though, so once your drive is pushing 60,000 hours, time to get suspicious again.
The oldest survivor in that particular study was a 320GB WD Caviar Blue that racked up over 154,000 hours, by the way. It'd be graduating high school next year.
SMART 187 is the one Backblaze acts on immediately
Uncorrectable errors mean the safety net already failed
SMART 187 counts reads that the drive's error correction couldn't fix, and this is the one where zero tolerance is the right approach.
Error correction catching and fixing problems is normal behavior on any drive, but an uncorrectable error means data actually came back wrong or not at all.
Backblaze schedules a drive for replacement the moment this value goes above zero, and when a company running masses of HDDs has a one-strike rule for a single attribute, I tend to follow suit.
SMART 188 catches a drive that's started hesitating
Command timeouts are the least reliable, but still worth a look
This one counts commands that took too long or got aborted entirely. It doesn't sound like anything special, but it can point to failing heads or electronics before any sector-level damage shows up, and that's pretty important in an HDD.
It's the least reliable attribute of the six, though. Backblaze tracks it, but not every drive even reports it, and a few isolated timeouts can just as easily come from a bad SATA cable or a failing PSU. So my rule here is calibrated accordingly: a missing 188 is fine, but a climbing one means something's going on.
SMART 197 is the closest thing to a countdown timer
Pending sectors, and why the median dead drive had just 13
Moving on to SMART 197, this one counts pending sectors, meaning spots on the platter the drive is struggling to read and has queued for reallocation. This is the closest thing to a countdown timer.
In the Secure Data Recovery study, the median dead drive had just 13 pending sectors when it failed, which does give me pause. The average was a massive 1,740, though. Some drives rack up thousands of these while others die almost clean. This is yet another proof that, unfortunately, drives can die even at 100% health.
The TL;DR here is that a rapidly climbing SMART 197 count is one of the strongest warnings a drive can give, but the study's data also showed that older drives often failed with counts near zero, which means a low number is less reassuring than it looks.
SMART 198 exists because manufacturers can't agree on anything
The backup counter for when 197 is being a little suspicious
Unsurprisingly, SMART 198 is essentially 197's sibling. It counts sectors the drive has fully given up on, but only those that are spotted during offline surface scans rather than everyday reads.
The reason to check both is that some manufacturers populate one attribute and not the other, so checking the pair gives you a clearer picture of what's up with your drive. The interpretation is the same as 197, too.
A clean SMART report isn't a promise
Circling back to that Secure Data Recovery study, over 30% of all drives failed with zero pending sectors, so unfortunately, no amount of SMART health checks will keep you safe. But doing so is just good maintenance, and it helps you pick up on some warning signs.
