One other big takeaway from the monster thread is that there is no single primary cause for the data being written to disk. That thread was titled on the premise that it was all due to high SWAP. But that is not the case, though maybe for some people. The TBW on my machines tended to high side but SWAP was not a factor for me. There are other theories that were supported by some. I once had 28TB written in the space of 31 hours, documented in this post.
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.
The lifespan of MacBook SSDs
- Thread starter Kyudoka
- Start date
- Sort by reaction score
One other big takeaway from the monster thread is that there is no single primary cause for the data being written to disk. That thread was titled on the premise that it was all due to high SWAP. But that is not the case, though maybe for some people. The TBW on my machines tended to high side but SWAP was not a factor for me. There are other theories that were supported by some. I once had 28TB written in the space of 31 hours, documented in this post.
Concrete numbers to calibrate: my M2 Ultra Studio 2TB (motion graphics primary, so a 250GB After Effects disk cache, ~150GB Redshift local cache, Premiere media cache doing its thing) shows 380TB written after 3 years and DriveDx reports 96% health left. If Samsung's 150TBW vendor floor were the real ceiling I'd already be at 0% and bricked, so clearly not.
For OP's 256GB / 8GB at 50GB/day, quick math: 18TB/year, so even at the pessimistic Samsung number you're 8+ years out. The multi-year torture-test threads people have run here have Apple silicon SSDs going well past 5-10x the vendor TBW.
Real failure mode nobody's raised: the SSD controller on Apple silicon lives on the M-series SoC, not on the NAND package, and the encryption keys are in Secure Enclave. If your SoC dies for unrelated reasons (bad reflow, water, surge past MagSafe protection), the NAND is fine but the data is unrecoverable because the keys walked with the SoC. That's the actual reason to keep a solid backup, not TBW wear.
For OP's 256GB / 8GB at 50GB/day, quick math: 18TB/year, so even at the pessimistic Samsung number you're 8+ years out. The multi-year torture-test threads people have run here have Apple silicon SSDs going well past 5-10x the vendor TBW.
Real failure mode nobody's raised: the SSD controller on Apple silicon lives on the M-series SoC, not on the NAND package, and the encryption keys are in Secure Enclave. If your SoC dies for unrelated reasons (bad reflow, water, surge past MagSafe protection), the NAND is fine but the data is unrecoverable because the keys walked with the SoC. That's the actual reason to keep a solid backup, not TBW wear.
chrono1081
macrumors G3
I wish I could upvote this 1000 times.
kkay2000
macrumors newbie
Bought a used Mac Pro 2013 ('Trash Can') with 64GB RAM and 1T SSD. SSDReporter shows 88% life remaining.
Note that all else being equal TBW scales with the size of the drive. So if a basic 250GB Samsung SSD has a 150 TBW rating, the same design in a 2TB capacity would likely have a 1200 TBW rating. You wouldn't be at 0% but you might be down to 60-70% (32% +/-) if a Samsung SSD was a model for your Mac's.
Like a car warranty, I wouldn't expect an SSD to fail immediately upon hitting its TBW rating. However, I haven't seen any real ones where the SSD go much past ~ 2x the rating.
I would not assume that one can routinely exceed a vendor TBW rating by 5-10x based on any endurance tests of Apple's internal SSD because Apple doesn't provide any TBW rating for internal SSD. We can only infer and more accurately speculate about their rating. They are certainly not 150 TBW for any recent SSD. Most likely in the range of 1000-3000x the rated capacity (e.g. I am sure yours is at least 2000 TBW and the 10000 TBW implied by the drive's health report is not unreasonable).
There are many reasons to keep a solid backup but that's also certainly a good one. Agree TBW is among the least significant.
On the flip side, I suspect loss and damage are the highest risk to data on laptops primarily used by high-school and college students.
There's a lot of different risk profiles out there but I can't think of any where no backup is a good idea. Well I don't generally backup the OS and a few other exceptions...
Analog Kid
macrumors G3
Back up your data because **** happens, but then you can pretty much stop worrying about it.
baconcat_8964
macrumors 6502
Data from smartctl shows mine SSD is currently at 87TB written, 100% available spare, and 3% used.
I used to worry about SSD lifespan on M1 Mac and macOS memory swap, but it seems like that's not a big deal after almost 6 years.
I used to worry about SSD lifespan on M1 Mac and macOS memory swap, but it seems like that's not a big deal after almost 6 years.
GMShadow
macrumors 68030
On a 2020 M1 13" Pro, 190TB written, 11% used. Barring a freak accident, in the practical lifespan of a Mac you're not going to see an issue.
BillyJoeJimBob
macrumors regular
It's an interesting problem to figure out. For those who believe that swap isn't an issue in the life of an SSD, then why would we ever need more than 32GB RAM for anything other than running local LLMs? Anyway, wear and tear from not being able to run the browser entirely in real RAM is why I use Firefox. I won't mention by name the only SSDs I've had fail within 2 years of purchase, but their initials are SanDisk (Western Digital) and Hynix. One of my oldest SSDs is a Samsung 840 EVO and it's fine. IOW, I think the quality of the chips may be in play here. What people are actually worried about is the fact that Apple welds everything together now, which makes it all disposable tech.
I don't think about spec'ing RAM and SSD life in the same context. I spec RAM based on how much I need to run all my applications/etc without swapping on the assumption that swapping is slow. It is much faster with today's SSD but SSD are still < 1/10th bandwidth and even worse on latency than RAM. What's the point of an M4 or M5 (or even an M1) if you are running at the speed of the SSD?
That doesn't mean I never let my system use swap once I have it. I've done real work on a Mac Mini 2018 w/8GB running SQL Server in a Docker Container, Tableau, Excel, Safari, iTunes, Mail, Calendar, and Notes all at the same time. Probably had PostgreSQL active too. That didn't fit in 8GB of course and so there was substantial swap use but it didn't continuously swap. iTunes never skipped a beat (literally). That was also Mojave.
The challenge these days is that it's hard to predict how much RAM an application actually needs. Modern versions of things like Office seem to need more RAM than ever with virtually no new features. They ask for as much RAM as they can get but then seem to run all the same when mostly swapped out. Browsers the worst (some better than others as you note).
However, I do avoid active swapping. If an application has a working set size greater than RAM (or actually about 75% of RAM accepting macOS's default parameters), it will actively swap. Active swapping can easily drive 10s GB of writes in a few minutes and that will wear down an SSD / shorten the system life if left unchecked. A system with a 256 or even 512GB internal drive would die in less than a year.
Howard2k
macrumors 604
DriveDX is layering a linear model on top of Apple's percentageLifeused SMART attribute, which is itself a vendor-estimated value not a directly-measured wear count — smartctl straight to the SMART table is closer to truth. The 150 TBW Samsung publishes is a warranty floor for the OEM supply contract, not a cliff; the old TechReport endurance runs and the Hardware.fr survey both had retail cells clearing well past 1 PB before UBER climbed measurably.
Data point on my end: kid's 2015 MBA 256GB, roughly five years of light use, DriveDX has been stuck at 96% for the last 18+ months while writes have very much kept happening. The reporting granularity is coarse enough that I wouldn't put much weight on small percentage swings.
Separate thing nobody talks about — on modern soldered Apple drives, the controller/DRAM buffer tends to age out before the NAND does. Seen a couple 2018 Airs come in with fully-readable data but PCIe that wouldn't renegotiate above Gen 1x1 anymore. Different failure mode than TBW, and not something DriveDX has any visibility into.
Data point on my end: kid's 2015 MBA 256GB, roughly five years of light use, DriveDX has been stuck at 96% for the last 18+ months while writes have very much kept happening. The reporting granularity is coarse enough that I wouldn't put much weight on small percentage swings.
Separate thing nobody talks about — on modern soldered Apple drives, the controller/DRAM buffer tends to age out before the NAND does. Seen a couple 2018 Airs come in with fully-readable data but PCIe that wouldn't renegotiate above Gen 1x1 anymore. Different failure mode than TBW, and not something DriveDX has any visibility into.
So you'd say that DriveDX is too pessimistic with its estimate, did I understand you correctly?
"swapping" and "active swapping" is a bit subtle and likely won't be clear. What I mean by active swapping is that the system is continuously reading and writing to swap. This is not easily seen in Activity Monitor (other than high Data written/sec in the Disk tab when you don't have any program open that you expect to do significant regular I/O).
However one can see "swapins" and "swapouts" in the top(1) program available from the Terminal. The man pages don't document those fields (at least on my current system) but they appear to be in the form of TOTAL (INCREMENTAL), where TOTAL is what happened since boot and INCREMENTAL is what happened since last refresh. If the INCREMENTAL value for swapouts is continuously (and significantly) positive, your system is "actively swapping". At that point I'd say you are working beyond its capacity. At which point maybe close applications / free up memory or try to work with less data at once. Or maybe it is not ideal for the task at hand.
On the other hand, just having a lot of "Swap used" in Activity Monitor isn't by itself terrible. My Mac Mini 2018 w/8GB running Mojave could keep ~ dozen large applications open without too much trouble. Yes Swap Used was relatively high as a % of total RAM but most of that was just detritus and application bloat that wasn't needed after it got swapped out. So everything ran close to 100% of normal and Tunes never missed a beat.