Hard drives have a shelf life too. There are exceptions. I still have a 2013 SSD. I had a 2019 HDD that died a year ago.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple's Regular Mac Base RAM Boosts Ended When Tim Cook Took Over
- Thread starter MacRumors
- Start date
- Sort by reaction score
(I'm in IT)
I have photo evidence that they fail, though wear may not have been the cause.
I have 2 different, barely used, dead ones on my desk right now. One is a standard consumer-class M.2 NVMe, and the other is a slightly premium OWC NVMe designed for Apple MacBooks from around 2013-2015 (though, maybe also up to MBA 2017, I can't remember which ones those take). These are both truly barely used, but both out of warranty. The computers they were pulled from are still functional, so it doesn't appear to be a board or PSU issue.
Since 2012, when Apple started using NAND on a proprietary blade form factor, I've seen out 10 or so failed drives in various machines. Since going to NAND directly on board around 2016, I've seen maybe 3-4 failed drives. Since Apple went with their own silicon in 2021, I've only personally seen 1 failed NAND (though that may have been power related).
For the above Mac storage, I can't say for sure if they were worn out, or failed for other reasons. Many MacBooks have an issue where power can jump on the rails and fry the SSDs. This is a known defect that Apple has yet to fix, and it's not related to wearing out of the drives.
For the 2 M.2 SSDs, they failed prematurely, and I'm not sure why. It could've been electrical, it could've been material and manufacturing defects, or it could've been that a certain portion of the drive was incorrectly written to too many times, we into premature failure. Could be something else considering as well.
Server class SSDs have to have a really solid lifespan, though Apple doesn't use that level of NAND (though, they sure charge us as if they were).
I think the bigger issue for drives is that they're not easily replaceable, and can't be pulled to recover data. As far as replacing due to wear, I think you would have to run in deep SMART report on many drives, every single day, then go back and look at the history once one or more fail. Unfortunately, I haven't seen anybody do a large statistical report on consumer drives.
All that being said, I don't think Apple ever purposely designs anything to fail prematurely, to increase sales, because that could hurt their brand, and leave them liable to a class action suit (and not necessarily out of the goodness of their hearts lol), but I do think they often make choices based on what they do and don't care about. Like, they want to design a specific feature or form factor, or design something to locking people into their ecosystem, services, and/or even their warranty and repair system, because they make a ton of money off of those things. And, if those choices cause a small number (in their case a few million devices) to die, due to premature failure of a certain part, or a defective design flaw on Apple's part, then they don't really care. Basically, I think it's more hubris or negligence, than malicious decisions, and what people like to call "planned obsolescence". I think what really happens with Apple's what looks like "planned obsolescence", and more, so the fact that they are dead said on steamrolling towards whatever goals they have in time, that they don't care what they leave behind, flattened, or destroy along the way.
Attachments
Everything fails, but is it “wearing out” from using it? Where is the evidence?
Everything fails eventually, but is it “wearing out” from using it?
Where is the evidence?
They would be, if an SSD was one component, they're they're grid of trillions of bits on several NAND cells, plus DRAM ICs, capacitors, controllers, traces, and so on. Something could be barely used from the perspective of the end user, but part of it could be worn out. Therefore, the statement isn't contradictory when you look at the whole of the device.
No, the SSD won't just hammer on a single cell over and over.
Wear leveling - Wikipedia
Where is the evidence that they don't wear from standard use? You can't assume one thing is right or wrong based on the lack of evidence. That's why what I wrote a bunch of information based on the evidence I've seen, which is not clinical or empirical, and did not a statement pointing to a conclusive cause. What I did say, though, is that I haven't seen any long-term studies on consumer level drives, only server class drives, but haven seen a good amount of SSDs fail.
I will add further information, my above statement is based on NVMe based SSD. I have absolutely seen SATA-based SSD fail due to overuse, including mSATA blade SSDs.
Now that I think about it, I do know for sure of one NVMe that got worn out. Someone made the mistake of buying a consumer level and NVMe drive to be used as a file server cache. The server was used heavily for video and design production. It lasted about 2 years before it died, and was not covered under warranty by Western Digital, due to being over the 800TB or 900TB write limit it was rated for (I think it was covered for 800-900TB over 5 years). I don't remember the exact numbers, as it was before COVID, so about 5 years ago.
Your comment don't make sense relative to the quoted text. Did you mean to reply to something else?No, the SSD won't just hammer on a single cell over and over.
Wear leveling - Wikipedia
en.wikipedia.org
Yes, they do. but have you done the math to find out WHEN the SSD is expected to wear out? I think the average Mac SSD will run longer than the lifespan of any owner.
The chip is good for about 1,200 billion writes if we assume competent wear leveling (source: Samsung's data sheet)
Lets assume the Mac runs 24x7 and does 10,000 writes per second. This means about 120,000,000 writes per year
So next we do some division, or do we even need to bother? The SSD likely will not last 10,000 years because something else over than writing to it will kill it first. Some passive parts like a ceramic decoupling cap, will fail first.
Last edited:
10,000 writes per second x 365.27 days (average days in a year, rounded to the hundredths place) = 315,593,280,000 writes a year.
In the USA, 1,200 billion is also called 1,200,000,000,000, if I'm not mistaken.
1,200,000,000,000 / 315,593,280,000 = 3.8023623317 years
That would be a just under 4 years.
Though, I know you've over estimated the writes per second, so maybe you meant 10,000 writes a day or an hour, but I'm not sure how you got to 120,000,000.
Last edited:
Please note that spec sheets and MTTF are statistical, calculated, estimations by the manufacturers—and at face value— do not represent real world performance, nor factor in manufacture defects, and environmental conditions. the only true way to know the lifespan of a drive is to measure the entire lifespan of a large sample of drives, in a range of environments.
Also note that I'm specifically not arguing for or against the lifespan of these drives, just merely pointing out all the information, as well as my personal experiences.
I do suspect that NAND cells, built within the last five years or so, would have a failure rate resembling something like:
80-90% never failing, due to the large majority of them never being heavily used.
10-20% failing due to manufactured defects and other environmental factors
Less than 10% failing due to heavy usage.
The numbers aren't a hard estimate, just a rough idea of how I could see them failing, based on the science of how they work, how wear leveling is implemented, how people use them, and just the overall history of human manufacturing.
Also note that I'm specifically not arguing for or against the lifespan of these drives, just merely pointing out all the information, as well as my personal experiences.
I do suspect that NAND cells, built within the last five years or so, would have a failure rate resembling something like:
80-90% never failing, due to the large majority of them never being heavily used.
10-20% failing due to manufactured defects and other environmental factors
Less than 10% failing due to heavy usage.
The numbers aren't a hard estimate, just a rough idea of how I could see them failing, based on the science of how they work, how wear leveling is implemented, how people use them, and just the overall history of human manufacturing.
No, I meant to reply to your "grid of trillions" and user perspective comment. It's not entirely clear what you mean by that, but it sounds like you're counting flash cells and suggesting that even though a user might not think they're writing a lot of data, beating up on one of those little cells could wear it out. That's not how these work.
It's entirely possible I misunderstood your point about trillions of bits and capacitors and whatnot...
I have a feeling you are not arguing in good faith and I will regret this but....
What is SSD Wear Level and How to Measure It (Detailed Guide)
If you keep important data on your SSD, read this. Learn what SSD wear level is, how to measure it, and how to prolong your SSD's lifespan.
news.macgasm.net
Does reading from an SSD decrease its lifetime?
SSD manufacturers advertise TBW (terabytes written). Why is "terabytes read" not part of the spec? Is that because reading from an SSD is virtually irrelevant for its lifetime? Added: For...
SSD Lifespans: How Long Can You Trust Your Solid-State Drive? - Datarecovery.com
One of the most common misconceptions about solid-state drives (SSDs) is that they’re capable of storing data indefinitely. All storage devices eventually fail, and unfortunately, SSDs are no exception. That doesn’t mean that they’re unreliable — SSDs offer much faster data access than hard...
datarecovery.com
SSD Lifespan: How Long Does an SSD Last?
They'll probably outlast your regular hard disk drives.
www.howtogeek.com
I have a feeling you are not arguing in good faith and I will regret this but....
What is SSD Wear Level and How to Measure It (Detailed Guide)
If you keep important data on your SSD, read this. Learn what SSD wear level is, how to measure it, and how to prolong your SSD's lifespan.
Does reading from an SSD decrease its lifetime?
SSD manufacturers advertise TBW (terabytes written). Why is "terabytes read" not part of the spec? Is that because reading from an SSD is virtually irrelevant for its lifetime? Added: For...superuser.com
SSD Lifespans: How Long Can You Trust Your Solid-State Drive? - Datarecovery.com
One of the most common misconceptions about solid-state drives (SSDs) is that they’re capable of storing data indefinitely. All storage devices eventually fail, and unfortunately, SSDs are no exception. That doesn’t mean that they’re unreliable — SSDs offer much faster data access than hard...
SSD Lifespan: How Long Does an SSD Last?
They'll probably outlast your regular hard disk drives.
It is very hard to wear out an SSD. The idea that Apple's memory sizes are going to kill your SSD is FUD until someone can show otherwise. With all the reports of this being a concern, I've seen nothing that looks like supporting documentation.
I have a feeling you are not arguing in good faith and I will regret this but....
What is SSD Wear Level and How to Measure It (Detailed Guide)
If you keep important data on your SSD, read this. Learn what SSD wear level is, how to measure it, and how to prolong your SSD's lifespan.
Does reading from an SSD decrease its lifetime?
SSD manufacturers advertise TBW (terabytes written). Why is "terabytes read" not part of the spec? Is that because reading from an SSD is virtually irrelevant for its lifetime? Added: For...
SSD Lifespans: How Long Can You Trust Your Solid-State Drive? - Datarecovery.com
One of the most common misconceptions about solid-state drives (SSDs) is that they’re capable of storing data indefinitely. All storage devices eventually fail, and unfortunately, SSDs are no exception. That doesn’t mean that they’re unreliable — SSDs offer much faster data access than hard...
SSD Lifespan: How Long Does an SSD Last?
They'll probably outlast your regular hard disk drives.
I know that it can happen, everything fails at some point. I’m talking about evidence that Macs have failed because of eight gigs of RAM.
I too have worked in IT supporting Apple products, and I’ve essentially never heard instances of SSD failure in Macs due to to much writing - despite the lame ram allocations.
Anyway, that’s enough, we can move along the next rumor now.
Last edited:
It is very hard to wear out an SSD. The idea that Apple's memory sizes are going to kill your SSD is FUD until someone can show otherwise. With all the reports of this being a concern, I've seen nothing that looks like supporting documentation.
Agreed, SSDs are robust.
I also agree that with a casual search I cannot find any studies showing side by side machines performing task X, one with 8gb and one with 16, 32 etc. trying to simulate the effects of swapping on drive lifespans.
However, if writing ages the drive then anything that increases writes accelerates aging, right? Same applies to a spinner, they were rated in hours or MTBF, the more the drive was in use the shorter the life span in days. Of course there was the arguement of power cycling wear but that is another conversation. This is why we have drives rated for datacenter or 24 hour use, like NAS or Server vs consumer.
I am not trying to by hyperbolic about this, I have never argued that 8g contributed to anything, but pretending that increasing writes to an SSD doesn't age the drive is silly. Circling back to drives being robust, we may just be producing drives that under normal use will outlive the CPUs usefulness, heavy writes or not, and that is a great thing!
All of this was triggered by someone noticing a bug in MacOS that caused it to write to disk at a higher rate than expected. That bug has been fixed, but the discussion has taken on a life of its own. Usually people support it by reporting the amount of swap used, which doesn't inform the discussion at all-- what matters is the swap rate, not the swap size.
There's swapping and there's caching, both of which are reported as "swap" by the memory manager. Most of the people who talk about swapping open up a bunch of Chrome tabs and look at the size of the swap file. That's an example of caching-- the system has decided to tuck away a pre-rendered web page so it's faster to redraw. It won't be written faster than a human changes tabs, which is extremely slowly.
Swapping is when you have a working data set that is larger than available RAM even when compressed. That will be written at machine speeds which could theoretically start to wear your drive, but frankly it would bog your system down to the point that you'd replace it out of frustration long before the SSD wore out. People doing work on datasets that large know it, and buy the RAM they need.
Memory intensive applications are very smart about how they use memory and swap. The reason you don't see stories about how different swapping behaviors wear out a drive is because it would be very, very hard to make happen and by the time you managed to kill the drive your results would be irrelevant because a whole new generation of technology would have arrived on the scene. That TechReport article took 18months of intensive, intentional, 24/7, drive writes. It would take 5-10 years of simulated "real world" virtual memory swapping to get a result.
Last edited:
I think the only real research I’ve seen in this area is that if you are *constantly* writing to the SSD for 20 years or something, it *might* have an impact.
Otherwise, the SSD might become inoperable sooner or later, but *not* because of writing to the silicon too many times. Said computer will be put in the recycling bin way before that because of failed battery, bad screen, slowness, no longer running current operating systems, etc., etc.
Does anybody here intend to be using their current 8gb laptop for another 19 years?
It’s FUD.
Otherwise, the SSD might become inoperable sooner or later, but *not* because of writing to the silicon too many times. Said computer will be put in the recycling bin way before that because of failed battery, bad screen, slowness, no longer running current operating systems, etc., etc.
Does anybody here intend to be using their current 8gb laptop for another 19 years?
It’s FUD.
Exactly, there are many additional components that are more fragile, subject to mechanical wear, get hotter, or handle more power that are much more likely to fail on any system that is 10+ years that has accumulated significant run time.