1

Most SSDs reviews focus their attention on speed and endurance (terabytes written) benchmarks.
However, searching a little bit more on the subject, I've found the following issues:

SSDs have very different failure modes from traditional magnetic hard drives. [...]
(for example, incomplete or failed writes due to sudden power failure can be more of a problem than with HDDs, and if a chip fails then all the data on it is lost, a scenario not applicable to magnetic drives). [...]
SSDs experience significantly higher rates of uncorrectable errors (which cause data loss) than do HDDs. [...]

Source: Wikipedia | Solid-state drive | SSD failure

Are power outages killing your SSDs?
[...] Of the 15 drives (10 different models, from five vendors), only one drive model, from one vendor, had no failures of any sort. One device failed completely (SSD #1), while one-third of SSD #3 became unusable due to metadata corruption. The other SSDs all exhibited various types of data corruption when they unexpectedly lost power, including the high-end enterprise SSDs with SLC NAND and supercapacitors. [...]
The implications of this research are significant. It suggests that SSDs, including enterprise SSDs, should not be trusted to behave in the proper fashion, or to be as robust as HDDs. [...]

Source: ExtremeTech

Investigating Power Outage Effects on Reliability of Solid-State Drives
[...] we show that failures in SSDs are not only due to volatile DRAM cache but also we observe similar failures in SSDs with disabled internal cache.

Source: Cornell Tech | arXiv

Therefore, in summary:

  • SSDs experience significantly higher catastrophic failure ratios due to power loss than do HDDs (ex: bricked device, metadata corruption, uncorrectable errors, ...).
  • SSDs, including enterprise SSDs, should not be trusted to be as robust as HDDs.

Questions

  1. Is my summary correct?
  2. Are desktop users really experiencing higher failure rates with SSDs than with HDDs?
  3. Should desktop users indeed be worried about bricked SSDs due to power loss?
Improve this question
8
  • 3
    #1 Yes, according to your sources (context is missing and extreme lab conditions do not reflect real world usage); #2 No, the opposite actually; #3 Not really.
    – ChanganAuto
    Mar 10, 2022 at 18:08
  • 1
    Also please note your 2nd source is almost 10 years old. SSD technology improved a lot since then. And yet this very outdated source seems to be then main anchor for your 3-in-1 question / concerns.
    – ChanganAuto
    Mar 10, 2022 at 18:22
  • 1
    @ChanganAuto > "#2 No, the opposite actually; #3 Not really." Mind to explain? Any sources?
    – Jeron Baffom
    Mar 10, 2022 at 19:07
  • 1
    As they say in academic circles, Wikipedia isn't a source ;) In this case it reflects the same outdated research. Please read the answer, nothing else to add.
    – ChanganAuto
    Mar 10, 2022 at 19:09
  • 1
    Scratch that... One more anecdotal evidence to add to the answers below: A State-owned and multinational company running tens of thousands SSD of various types, connections and form factors (pretty much all of them) with an extremely negligible failure rate, way below what was planned for years ago. OTOH, many HDDs in servers, supposedly the best of the best, failed catastrophically and keep failing. Not a problem, just the cost, because RAID and backups, of course.
    – ChanganAuto
    Mar 10, 2022 at 19:18

3 Answers 3

Reset to default
3

The best source of data is the Backblaze Hard Drive Data and Stats for their stock of thousands of hard-working HDDs and SSDs.

They have dedicated one article to this question: Are SSDs Really More Reliable Than Hard Drives?.

When comparing HDDs and SSDs of the same age, the results are:

enter image description here

SSDs win by a very small difference. In fact, both drive types are within the other’s 95% confidence interval window after 14 months.

If we are looking instead at data dating from several years, the picture is:

enter image description here

The article sums it up:

As the graph shows, beginning in 2018, the HDD boot drive failure rate accelerated. This continued in 2019 and 2020 before leveling off in 2021 (so far). To state the obvious, as the age of the HDD boot drive fleet increased, so did the failure rate.

It is still too early to give a conclusion, but the two curves look very similar. However, for every single age the SSDs win by a small margin.

For the few years that Backblaze was using SSDs, they may seem slightly more robust than HDDs, but it's still too early to say what the results will be in the coming years.

Improve this answer
7
  • I guess you have flawlessly answered the following: In a controlled environment without power failures, both SSDs and HDDs have similar failure rates. +1
    – Jeron Baffom
    Mar 10, 2022 at 21:39
  • However, what if you compare them in a standard home setup? ie: a couple of power failures during their lifetime. Does the aforementioned conclusion remain the same?
    – Jeron Baffom
    Mar 10, 2022 at 21:43
  • I don't think there are any statistics about this. All I can say that my own experience is the same as in the other answers: I have had HDD failures, but not yet any SDD failures.
    – harrymc
    Mar 10, 2022 at 22:09
  • > "I have had HDD failures, but not yet any SDD failures." You never experienced any SDD failure? Mind to explain that a little bit better? How many SDDs have you tried? For how many years? Do you have an UPS?
    – Jeron Baffom
    Mar 10, 2022 at 23:05
  • BTW, I've just read carefully the article you've posted from BlackBlaze: "Are SSDs Really More Reliable Than Hard Drives?". Indeed very good (+1). Just as a reminder: The comparison was made for boot drives, ie, low workload. Now I'm wondering: What would happen for high workload?
    – Jeron Baffom
    Mar 10, 2022 at 23:08
3

Before this gets closed as opinion-based, my $0.02…

They all work until the day they don't.
You always need a backup. After that, who really cares?
It takes a couple of hours to be back up & running once you swap a drive - so long as it's swappable.

That would almost be my biggest concern these days, soldered memory/drives. I'm about to drop 4 grand on a machine with no user-replaceable parts…

Otherwise, HDs for long term storage & backups [plus off-site, belt & braces], SSDs for speed & often-used data. 8TB SSD prices still make your eyes water - so big drives, HD for now. That will eventually change.

Anecdotally, I've only lost one SSD since I first switched all my boot drives to them maybe 6 or 7 years ago. It was a cheap one. I was back up in 2 hours.
I once lost 3 HDD boot drives in a year. Again, back up in 2 hours.
My oldest 1TB Samsung is now 8 years old & still my primary drive on my main work machine. My HDs are all newer because I sacked smaller ones to fit larger ones, not because they failed.
I am not going to extrapolate from such a small sample set. Nor should you.

Improve this answer
3
  • 1
    +1 Exactly!....
    – ChanganAuto
    Mar 10, 2022 at 18:09
  • 1
    "You always need a backup." +1 The only SSD I've had fail so far was an OCZ Synapse Cache from many years ago. It was great for the year-and-a-day it worked... (Actual time may be inaccurate, but just after the warranty expired.)
    – Andrew Morton
    Mar 10, 2022 at 18:46
  • > "Before this gets closed as opinion-based" Questions #1 and #2 are objective questions and not mainly opinion-based.
    – Jeron Baffom
    Mar 10, 2022 at 19:10
1

I have a ton of SSD and maybe 1-2 of them failed due to age.

I have had power outages, and not a single SSD was damaged.

Now on the other hand I had piles of hard drives that failed or we unusable due to bad sectors.

I just had to dispose of 22 of them, and had to pay to have them securely erased.

I have had dozens of hard drives with motor failures, which SSD don't have any so they can't fail in this manner.

I have had thousand upon thousands of bad sectors on those hard drives. For which SMART NEVER remapped ANY significant number of them. Drives used up their re-mapable sectors, and never warned me of their impending failure. I had many many drives that stubbornly refuse to remap ANY bad sectors at all.

The SSD does actually have a ton of spare sectors and does in reality remap them.

Even worse I have had hard drive drop to less than 2/3 of their new performance because it was basically ECC'ing every single sector on the drive. In certain areas of the drive performance was less than 1mb/s.

Every moving part on a hard drive is another potential point of failure which SSD(s) don't have.

In my long experience with numerous hard drives, I have never found them to be reliable long term.

For me hard drives are nothing but tears (from lost data),pain,slowness, and frustration. Although now everything is properly backed up multiple ways.

Improve this answer
11
  • 1
    @JeronBaffom Again, Wikipedia mentions oudated research from many years ago about the lifespan and failure rate of drives a of which the newest ones are from 2015 and the oldest probably predate 2010. And all are from heavy usage (Google's). The relevance of this findings must be put into its temporal context and shouldn't be extrapolated to the current generation of SSDs of which we have nothing to compare them with yet, no similar study has been conducted since. And the relevance of those finding for desktop users is, being nice, none at all.
    – ChanganAuto
    Mar 10, 2022 at 19:30
  • 2
    SSD do wear out individual sectors faster when data is written to them. However, they have thousands of spare sectors, and most hard drives have 80-300. The SSD firmware automatically handles bad sectors, where as I would give hdd drives a 10% grade for handling bad sectors. Also a hard drive usually has <16mb of cache and ssd have 128mb maybe more so sometimes changes happen in DRAM before they are written to cache. Saving many actual writes from happening. So while wear level afters individual sectors more the SDD has so many TB's of writes that most people won't experience failure.
    – cybernard
    Mar 10, 2022 at 19:30
  • 1
    I also agree with @ChanganAuto, because before wear-leveling was a thing there were some dark times for SSD. The couple SSD that I had fail on me pre-date wear leveling. Modern SSD are leagues better, just don't buy the dram less kind as you will suffer a hit to performance.
    – cybernard
    Mar 10, 2022 at 19:33
  • 1
    @cybernard > "hard drive usually has <16mb of cache and ssd have 128mb maybe more" I'm not so sure whether write caching is that advantageous. Take a look at the warning of my drive settings: "Performance is increased by enabling write caching, but leaves the system susceptible to data loss in the event of a power failure."
    – Jeron Baffom
    Mar 10, 2022 at 19:46
  • 2
    @JeronBaffom Both can lose data, under rare conditions. However, enterprise SSD all have extra capacitors so if it takes 1 second to write all the data from the cache to storage they have 2 seconds worth of capacitor or etc. This is impractical for a hard drive because of the comparatively large power needs of the motor. Many times SSD either have enough residual power, or can drain residual power from capacitors in the power supply to finish writing. Yes, data loss can happen but its rare. Even then its only the fraction of the cache it didn't write.
    – cybernard
    Mar 10, 2022 at 20:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.