How long do HDDs actually last

When I look on google, they say they last 3-5 years.

But I have a laptop that is almost 20 years old and still works. I just pulled a laptop out of a box that is 13 years old and it still works. I have external HDDs that I bought ten plus years ago that still work.

I recently bought a large external HDD to back up all my files in case any of the smaller external HDDs end up breaking. But I’m wondering if this new external HDD will last 3 years or 20. I can’t find a good answer.

I’ve read SSDs can last 10-100 years. But I have a flash drive that was almost 10 years old and it was already giving me problems.

My PowerMac 7100/80’s original 500MB HD died…I’m going to say circa 1999. Quite young, as I’d only acquired the machine around 1996. I replaced it with a 2.1 GB and later added a 3.2 GB HD since it didn’t have a CDROM drive but had the SCSI and power ribbon cables for one.

That computer was on continuously until 2006. I still boot it now and then, sometimes for days at a time, running legacy software or making use of legacy hardware (e.g., converting a friend’s twelve boxes of floppy diskettes into disk images). Both disks are in good condition, and doing better than the machine’s aging cooling fans.

Another computer of mine, the PowerBook G4 I use as a FileMaker Server, has run nonstop from 2009 to present day. The hard drive it came with died on me about three years ago and has been replaced with an SSD.

Sometimes Winchester HHDs last for 3 years, sometimes they last for 30. I think there are fine differences in balance and friction and when they’re spot-on perfect there’s very little wear even when they stay spinned up nonstop. SSDs don’t have moving parts; it’s my understanding that their wear and tear factors are all about writes and erases. Boot from one and run it nonstop and it may last a lifetime; try to use it as an archive disk and erase it entirely and write a full disk copy of some other disk to it every night and it may die on you in just a few years.

https://www.backblaze.com/cloud-storage/resources/hard-drive-test-data

This is a relatively independent, large dataset on HDD longevity.

The Backblaze link is good, but it illustrates something I was going to point out. It’s not the years of life as much as the usage. The moving parts wear out over time. The bearings wear down or the head start hitting the platter, and physical damage occurs.

Disks die for a range of reasons.
Keeping them spinning is a good thing in general. A very common failure is a failure to spin up again.

Any Winchester technology drive - which is basically all of them - rely on flying the heads over the surface. This is pretty very robust, and can even survive reasonably serious shocks - although there are clearly limits. But success relies on lack of contamination inside the drive. Even the merest speck can cause a head crash and kill the drive.

Other failures come about because of degradation of the drive electronics. These can be caused by all manner of failures in head drive or read amplification. Electronics don’t last forever, and over time all sorts of components quietly slip out of tolerance. Eventually the controller processors will succumb to diffusion effects that will stop the transistors working. That takes many decades, but will happen.

Thermal cycling is a killer. That is true of just about everything. Things creep and fatigue, and there have been evil failure modes where things like adhesives and binders fail after so many thermal cycles.

One idea for very long lived drives has been to keep them spinning at a slow speed and keep them very cool. I’m not sure how much this is done now.

I have worked with systems where the disks have lived for over a decade. But not all of them. I would never rely on a single disk living for any amount of time. The number of new disks that turn out to have design or manufacturing flaws that only come to light some months after release should give anyone pause.

Flash drives are a problem because there is a maximum number of writes each internal storage element can cope with before it starts to fail. All flash drives use a wear levelling system to even out the number of writes, and keep a reserve of pages to extend the life of the drive. The higher spec drives have more reserve pages, and often use a more robust (and less space efficient) layout of the storage elements. You get what you pay for. Flash drives for a system disk are good in some ways. But if the system pages a lot, or writes and cycles through lots of logs, you might find the drive failing much sooner than you hope for.

Don’t reply on a single drive for anything important. Certainly not backups. My usual view is to treat the computers one uses as a cache of data. Always assume that one day your laptop will burst into flames, or get run over by a truck. Backups need to be built on resilient storage. Minimally able to sustain the catastrophic failure of at least one disk. So minimally a mirror, and preferably using something that is able to cope with all manner of failure modes - which trivial mirroring can’t.

Yep. 3-2-1 for anything you care about keeping.
3 copies
2 different storage media types
1 copy physically separated from the others (offsite, in the cloud, in a separate building, or at very least not in the same box in the same room).

If one of the copies is in the cloud, that might fulfil more than one objective, or indeed, all of them if the cloud service provider is doing their own backups, which they should be.

IMO unless you’re dealing with multiple dozen terabytes of dynamic data, doing your own backing up is sooo 1990s.

My paltry half-TB of largely archival data is synced to one of the big cloud providers. It’s a certainty that Google, Microsoft, Amazon, etc., have far more robust storage and backup and redundancy and security and every other “ility” that matters than you do. Why go to a lot of trouble to manually do badly something that can be done flawlessly automatically by hired pros?

Internet speed, especially upload speed, restore time, security, and robustness*.

*Cloud-based backup services generally do not backup your entire computer. Backblaze for instance excludes Applications and other system folders. There’s reasons for that, but disk-based backup software can provide bare-metal or nearly one-step restores while cloud (at least personal cloud services) can’t.

Did you ever actually do that?

I have and it made me stop backing up whole systems.
Just not worth the bother.

I don’t need my computer backed up, I need my data.

This.

At least under Windows, attempting a byte for byte restore onto different hardware is asking for eternal glitches or a system that won’t even start.

The only time I’ve ever needed a restore it was after the laptop bricked. Buy a much newer more capable one with a fresh OS on it, download and install the half dozen specialist apps I use beyond Office, take the opportunity to upgrade that to latest version too, then connect to my cloud account and in an hour-ish I’m back in business.

i currently have about 5TB of data I want to back up. I was under the impression it would cost $10 a month or more to store that much data on the cloud. By comparison, external hard drives are running about $15/TB on amazon right now. So financially it never made sense for me to use the cloud as a backup storage option because I could buy an external HDD for about the same price as 6 months of cloud storage.

Aren’t HDDs and flash memory dropping in cost by 50% per TB every 3-5 years or so? I feel like it might make more economic sense to just buy two large HDDs and store the data on both of them rather than spending money on cloud storage.

Having said that, the bulk of my stored data is movies. I would miss them, but I don’t ‘need’ them.

The files I truly truly want to save are much smaller and are GBs of data, not TBs. I have multiple redundancies for the files I truly want to save. Those are backed up among several smaller HDDs, as well as stored on my laptop and on a flash drive. Some are also stored on google drive.

For the ~10GB of data I truly want to save I think I have them stored on 4 or 5 different platforms at this point.

Ten bucks per month is about right for Backblaze (previously mentioned) if you pay one month at a time. It’s $99 for a full year if you pay all at once or $189 for two years paid all at once. Yes, an external drive is cheaper but the advantage of a service like Backblaze (full disclosure: I’m an existing customer) is that it’s continuous and the backup is offsite. An external drive only has the data from your last backup and the backup drive is onsite unless you go to some trouble to take it to work or someone else’s house.

I actually do both; I use Backblaze and at month-end do local backups to portable (2.5" USB-powered) external drives that are stored in a fireproof safe in the other room.

Hard drives life expectancy is some sort of survival function. Every drive has an x% chance of dying every year, and the percentage is not constant. Generally it will go up as they age. The linked Backblaze numbers are probably the best real world estimates for the particular drive models they use.

The only way to think about hard drives is that they’re already dead, you just don’t know it yet. You have n-1 copies of your data. If it only exists on one hard drive, then you have zero copies of your data, hence the rule above about keeping three copies, which really means you have two, which means you have one backup.

I agree about external drives being cheap compared to cloud storage, but the most important part of cloud backups is moving the data to another location. Backups that will perish in the same fire or flood as the working data are not much of a backup. You can be creative about this, though. My “cloud” backup is an external drive I own attached to my work computer in the office.

All of this applies just as much to solid state drives as to spinning drives. The failure modes are different, but what do you care if the lubricant evaporated or if the flash failed prematurely? The data on the drive is still gone. I’ve had SSDs die long before they get anywhere close to their rated write cycles.

Yep. And a technique that SOMETIMES works when one won’t start is to take it out, set it on a hard, flat surface, and spin it around its central axis a few times, then reinstall and try again.

If the startup failure was due to stiction, this can save the day.

And then, obviously, get your data the hell off it ASAP!

I once had a HDD that failed and tried the freezer trick. (Stick the bare drive in the freezer for a while to unstick whatever is stuck.) It worked, at least long enough to get the data off of it.

If you think about what’s going on in a HDD, it’s amazing that they work as well as they do. A typical drive is one that’s 5400RPM, meaning that the platter spins round 5,400 times a minute, while a read/write head is moving back and forth a tiny, tiny amount of space above it. And it can do this for years, meaning millions or billions of rotations with that drive head moving around above it.

3-5 years is a good rule of thumb for computer obsolescence in general. That doesn’t mean a component won’t work after that amount of time. It’s just more likely to fail, and depending on the component it may limit what software you can run.

Some components like the battery on a motherboard (called a CMOS battery or CR-2032) is not expected to last more than 5 years, and rechargeable batteries will also likely have very limited capacity by that point. So there is a definite “expiration date” on some items.

I’ve had HDDs last much longer than 5 years though.

I currently have a subscription with Microsoft 365 that not only gives me access to Office programs but it backs up data I really care about in designated places. The cool thing is that I can share that data between multiple devices and of course it’s being backed up as well. But I don’t have anywhere near 5 TB to deal with.

While this is true, this isn’t the use case for full image backups. Besides the obvious ability to do a restore if the system HDD or SSD goes south, I’ve frequently used image backups to swap out system drives for larger or better ones, such as replacing an HDD with an SSD.

It’s cool if that works for you (and this is a model that works for many others, too, typically casual home users). But for business and more serious users, that “hour-ish” can turn to months, especially if upgrading to the latest and greatest involves a new OS.

Real-life example. I used Windows XP for so many years that I’d accumulated quite a suite of applications on it, some major and well-known, some niche. Shortly after Windows 7 came out, I installed it on a spare laptop solely to use as a migration testbed. Long story short, between applications that didn’t work properly or didn’t work at all on W7, and some that couldn’t be upgraded but had to be replaced by something different, the elapsed time to migrate my environment from XP to Windows 7 was approximately three months. There’s a reason that Microsoft provided a Windows XP virtual machine with Windows 7 Professional.

There’s not much to add to the good points that others have already made, except to agree that longevity is very much dependent on factors like operating temperature and other environmental conditions, number of spin-up cycles, build quality, and other factors. But “3-5 years” is just silly. Right now, on what I think of as my “new” desktop computer, the secondary HDD is reporting a power-on time of 43,208 hours, which is equivalent to almost five years of continuous operation, and I expect it to last much, much longer than that.

The traditional metric for the longevity (well, technically, the reliability) for electronic equipment is MTBF (Mean Time Between Failures). The measure can be problematic because until a device has been in the field for many years, it’s just a calculated theoretical value. So drive manufacturers have moved to a different metric, Annualized Failure Rate (AFR) because it’s a more practical measure that helps businesses determine how many spares to stock. But it doesn’t tell you how long, statistically, one given drive is likely to last.

AFRs tend to be sub-1% sometimes excepting the newest high-density drives that may exceed 1%. As for MTBF, although Seagate has abandoned that metric, they do cite calculated HDD MTBFs of between 300,000 to over a million hours, but also say that real-life field MTBFs tend to run between 50% to 60% of those numbers.

Taking that at face value, you get a field MTBF under proper environmental conditions of at least 150,000 hours, or about 17 years. How realistic this is I can’t say, but it does tend to match my own experience. I have HDDs quite a bit older than that (though they’ve not been in anything close to continuous use).

I currently have four computers, two of which are desktops with at least one HDD, and dozens of external HDDs, and have never had a catastrophic HDD failure (but of course backups are essential because it can always happen).

There’s also the issue that SSD’s have a finite number of rewrites before they stop working. I’ve heard various quotes between 20 and 100-plus years of common use. My 2011 MacBook with minimal use still works (after replacing the battery).
I find this:

On average, SSDs last between 5 to 7 years under normal usage conditions. However, this can vary based on the quality of the SSD, the intensity of usage and the factors mentioned above.

and also this:

As for hard drives, the oil that lubricates the platter bearings will pool and the bearings will seize up if the drive is not spun up once every so often.

OTOH, I have a PC that’s pushing 24 years old - I fire it up every so often, the disk seems to be good. Indications are the platter magnetization is good for a long time.

I agree that I prefer to do my own backups, which I try to remember every month or two - nothing is critical. I have several external hard drives. I back up photos, music, ebooks, My Documents, and the PST files for email. I also have a folder of downloaded installs for assorted programs (install keys saved in Documents). I also have a few data folders for other projects. I have an external drive where I packrat downloaded video, but for the expense of the extra disk space, for now one copy on an external drive is sufficient.

The risk is of course that someday some software, the verification process for the product keys online will no longer work. Or as I found once, I changed email and my old provider disabled my old email, so verification emails no longer worked. Fortunately, I found the original keys in an old document. After all, there may come a time when Yahoo or Google decides free email is no longer a worthwhile thing, like my old ISP did.

Just to add a data point to this. As I mentioned, I have dozens of external HDDs. The more recent ones (about six, a few years old or newer) are Western Digital 2.5" (“mobile”) drives, the most recent of which are amazingly thin and light. But I have quite a few older ones that are relatively large and heavy 3.5" drives (mostly a mix of Western Digital and Seagate). Most of those haven’t been powered on in three to five years or more.

Based on your plausible suggestion, I powered them all on one by one, looked at the files, and let them run awhile. All were fine. So again, the moral of the story is that modern (or even not-so-modern) HDDs are pretty reliable. And also again, they can fail at any time (and so can SSDs) so keep that in mind, too!