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.
ssd swap - high usage of Terabytes Written
- Thread starter Forti
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
According to DriveDx 1.10.1 my M1 Mac mini 16GB/2TB has written 3.8 TB in 7 weeks and read 7.7 TB.
If true and assuming 6,000 TBW allowed for the internal SSD,
we find 6000/(3.8*52/7) = ~212.6 years use expectancy at the current rate.
So far ¼ of the SSD capacity has been used.
At this point I see no need to disable Spotlight to diminish the usefulness of my M1 computer.
For comparison these are the TB numbers for my late 2013 Mac Pro 64GB/1TB:
TBW = 79.1 TB
TBR = 219.8 TB
BTW I keep this "trashcan" computer on macOS Mojave.
It's still going strong, including with 32-bit apps and older now obsolete peripherals.
If true and assuming 6,000 TBW allowed for the internal SSD,
we find 6000/(3.8*52/7) = ~212.6 years use expectancy at the current rate.
So far ¼ of the SSD capacity has been used.
At this point I see no need to disable Spotlight to diminish the usefulness of my M1 computer.
For comparison these are the TB numbers for my late 2013 Mac Pro 64GB/1TB:
TBW = 79.1 TB
TBR = 219.8 TB
BTW I keep this "trashcan" computer on macOS Mojave.
It's still going strong, including with 32-bit apps and older now obsolete peripherals.
Last edited:
Since I'm monitoring this problem since the moment I got my MBA 16Gb/1Tb, I can share some observations:
1) My "use case" is work-related and not extreme at all: some browsing, productivity tools, Office 365, Teams for calls and some lightweight development tooling. I try to observe memory usage and not to stress virtual memory, e.g. tend to use a single browser instance only, close apps, windows, etc. to avoid excess swapping.
2) To date I'm at almost 50Tb written during just 6 weeks of everyday use, and that is clearly not acceptable by any measure, even if drive could survive it for few more years - it's an overhead also from the performance standpoint.
3) Spotlight, Time Machine, iCloud and photoanalysisd are not the real culprit for the high TBW, since I have them turned off for at least a week, and still observe abnormally high disk writing rate (0.3-0.6 TB per day on the light load) almost exclusively from kernel_task. The total TBW is lower than before though, but not dramatically, which again points to the OS level VM management issue.
4) I migrated from MBP 16 2019 64Gb which I used for around 1 year, and there TBW was in the low single digits for exactly the same usage pattern and software configuration.
My primary suspicion is that Big Sur introduced a bug in the VM management algorithm, which is somehow more pronounced on the Silicon unified memory architecture, but also it looks like it is not exclusive to it and some Intel machines are affected as well. What also worth checking is the regression after the migration from the higher memory configuration - I personally don't have time test that, but wouldn't be too shocked to learn that it's the case.
Regardless of the reason, it must be fixable by the OS software patch from Apple.
1) My "use case" is work-related and not extreme at all: some browsing, productivity tools, Office 365, Teams for calls and some lightweight development tooling. I try to observe memory usage and not to stress virtual memory, e.g. tend to use a single browser instance only, close apps, windows, etc. to avoid excess swapping.
2) To date I'm at almost 50Tb written during just 6 weeks of everyday use, and that is clearly not acceptable by any measure, even if drive could survive it for few more years - it's an overhead also from the performance standpoint.
3) Spotlight, Time Machine, iCloud and photoanalysisd are not the real culprit for the high TBW, since I have them turned off for at least a week, and still observe abnormally high disk writing rate (0.3-0.6 TB per day on the light load) almost exclusively from kernel_task. The total TBW is lower than before though, but not dramatically, which again points to the OS level VM management issue.
4) I migrated from MBP 16 2019 64Gb which I used for around 1 year, and there TBW was in the low single digits for exactly the same usage pattern and software configuration.
My primary suspicion is that Big Sur introduced a bug in the VM management algorithm, which is somehow more pronounced on the Silicon unified memory architecture, but also it looks like it is not exclusive to it and some Intel machines are affected as well. What also worth checking is the regression after the migration from the higher memory configuration - I personally don't have time test that, but wouldn't be too shocked to learn that it's the case.
Regardless of the reason, it must be fixable by the OS software patch from Apple.
Ranter? You can take your holier than thou attitude somewhere else.
If you bought a car and it used all the tires unevenly and in 3 months used half the tread, you would be screaming.
Now tell us all why you wouldn't be "ranting."
Good news. Thanks for your in depth check. We hope that Apple will address this in a near future update.
That is about 2.56 TBW/power on hour. The fact that the output is edited is also odd but I just can't believe 2.56 TB/hour. For reference I'm at about 52 GB/power on hour. That machine was writing at ~50x more than mine? Odd to say the least.
I got 77 GBW / power-on-hour. There were people, in this topic, who got multiple more. On twitter you can find a few with even 1-2 TBW/hour.
Like I said before - I notice, that some people even with tough tasks (e.g. photo/video editing) have low TBW while others with low-tasks (mostly web browsing) have more TBW/h than mine Mac mini.
In my opinion - this is kinda proof that there is an issue.
I definitely believe that there is an issue. I just see something like 2.5 TBW/power-on-hour and I wonder if someone set it up to fail. Another big worry in the reported case is that the drive didn't get the expected 1.5 PBW but only about a third of that. All around odd.
The SSD in my trashcan pro died after about 4 years of use (2017). I am convinced that it was due to the temperatures because of its location on top of one of the GPUs - error in AHT was "temperature sensor SSD". As you say, it showed plenty of warnings about dying. Started off with occasional kernel panics, which became more frequent. Then sometimes it would show ? at boot and things got progressively worse from there.
First case of a failed M1 SSD.
When will apple release an official statement already ...
Curious though where are you getting 6000 TBW for the internal SSD, when drives like the Samsung 980 pro 1 TB are rated for 600 TBW. What NAND is being used in Apple's drives? Does anyone know?According to DriveDx 1.10.1 my M1 Mac mini 16GB/2TB has written 3.8 TB in 7 weeks and read 7.7 TB.
If true and assuming 6,000 TBW allowed for the internal SSD,
we find 6000/(3.8*52/7) = ~212.6 years use expectancy at the current rate.
So far ¼ of the SSD capacity has been used.
At this point I see no need to disable Spotlight to diminish the usefulness of my M1 computer.
For comparison these are the TB numbers for my late 2013 Mac Pro 64GB/1TB:
TBW = 79.1 TB
TBR = 219.8 TB
BTW I keep this "trashcan" computer on macOS Mojave.
It's still going strong, including with 32-bit apps and older now obsolete peripherals.
Based on the tweet above seems like 600 TBW is indeed the number.
Don't take this guy too seriously - he starts an article on a possible serious hardware issue with the line "But my response has always been the same: would Apple really sell a laptop that could potentially fall into disrepair before its standard warranty expires?".
No-one here expects their machine to fail in less than 12 months. But every computer I've ever owned has lasted years - even over 10 years.
Is seems unlikely to me my new Mac mini M1 will do that with its current burn-rate and non-replaceable components.
Oh, and I am neither "undertaking extremely intensive work" nor "consistently running benchmarks to stress test the hardware" as he concludes.
Samsung 980 pro 2TB is 1200 TBW. So we are still quite worlds' apart from 6000 TBW .
The tweet above suggests 600 TBW for the 512 GB Apple SSD. Generally speaking, that would mean the Apple 1 TB will have 1200 TBW and the 2 TB will have 2400 TBW.
Correct, taken from the article quoted earlier in this thread. Many thanks.
The 3000 cycles number was quoted for Toshiba SSDs Apple apparently has been using.
Can it be, as suspected earlier in this thread, that the tools are showing erroneous info due to new architecture?
I mean, 2.56 TBW/Hour, is there even capacity in the HW for that amount of data?
Last edited:
The TOOLS are just showing the SMART Data from the drive itself. If you don't believe the tools, use smartctl
command line.
I uploaded all that info from smartctl to my open cases for both computers. I get tired of Apple knotheads telling me that they can't look at a third-party tool.
I'm wondering why Apple hires so many tools.
That's true. I only use smartctl my self, both in *nix and now on Apple. It was just a thought.
this subject seems to have gone by the way side, in terms of front page news / Apple giving any sort of response or descriptor of an update coming that might fix it / it being a non issue
what gives
what gives