ssd swap - high usage of Terabytes Written

[cal6n]

Just about everything I have read on this issue - a lot - leaves me with more questions, never answers! If we accept that all components, including SSDs, can fail, what are we left to be concerned about? A bunch of reports from people concerned about what Smart Monitor is telling them; these are not to be dismissed because most of those people appear to be well-informed and very far from stupid. But if I look at what Smart Monitor reports on my two in-service Macs I am immediately struck my some figures that are very obviously wrong. Because I am sat at my iMac just now I show its Smart Monitor report produced in the last few minutes.
Power On Hours 491? Rubbish, this machine is switched on eight hours per day seven days per week, and has been since 01/06/2019.
Power Cycles 4538? No siree.
Unsafe Shutdowns 37? Get outta here. I use a UPS precisely to prevent unsafe shutdowns.

So why should I believe anything Smart Monitor tells me? This is on an Intel Machine and it's producing gibberish, how much less is it to be trusted on Apple Silicon? As it happens, the most believable figure is what it gives for TBW, but that doesn't concern me.

I certainly believe M1 machines write a lot of data to SSD, probably more than their Intel predecessors, but does that mean they're doomed to early (within 10 years) failure? I rather think my M1 will be good for that long haul, and that you can buy with a reasonable level of confidence. After all, if there's a software problem than Apple will fix it (obviously without admitting there's an issue!) and if there's a hardware problem (affecting a small number or users...) Apple will similarly offer a fix.

Don't worry, be happy, give Apple your money.

View attachment 1743521

The "Power Cycles", "Power On Hours", and "Unsafe Shutdowns" do not refer to your Mac's behaviour. They refer only to the SSD.

 

[osplo]

So what can I do? I have been postponing purchasing a MBA M1 for a very, very long time because of this SSD issue.

My two cents:

1) some high writing activity reported here turned out to be normal (such as 8GB owners opening 500+ tabs at once)
2) some users here have really high SSD writing even after disabling time machine and spotlight and more

I think that the percentage of people affected by (2) is relatively small considering the thousands (hundreds of thousands?) of M1 machines sold by now. It could be a Big Sur bug, even a hardware bug, but it is clear that it does not affect everyone.

I purchased an MBA M1/16GB (even after reading this thread) and I do not think it is affected. I started from a clean install and I am doing moderate office use so far (waiting for Adobe to migrate everything to M1). Super snappy, super happy with it.

I do want a fix and I do want Apple to take a serious look at this matter.

But I would not wait to buy an M1 machine for this. You can always return it after two weeks if you are one of the unlucky winners.

 

[pcharlesmorphy]

But that looks like it is not Big Sur's issue but Word's. It;s like the joke Mac users had back in the late 1980s.

User: I have a problem with my Mac.
IT: Oh what is it?
User: Well I am using Microsoft..."
IT: Stop right there. That is your problem...using Microsoft anything.

LibreOffice does much the same thing as Office and can import Word documents. Does this file have the same memory issues in LibreOffice as in Word? If so then things shift toward the OS but if not then the problem is Microsoft not knowing how to program not Apple's OS having a bug.

I didn't know this joke from Mac users about Microsoft Software

It is possible, that it is a Microsoft Word problem with memory management, and has to be optimised.

I still think there is a problem with Big Sur.

Without going any further, 20 minutes ago, I had an anomalous behaviour.

I had iMovie open, where I had put a 14-second video of low quality, to add sound to it.

While I was looking for an mp3 file, after approximately 20-25 minutes, I got an error message saying that I was out of memory and I closed the programs.

I had iMovie, Safari with 4 tabs, Whatsapp and Telegram.
So kernel_task had written 58Gb!!!!

It is very clear that Big Sur has a problem, or it is a problem with the architecture, because it makes no sense that with 16Gb of RAM, and in those conditions this has happened.

 

[wirtandi]

Microsoft Word and Excel have already been optimized, right?

 

[Fomalhaut]

How can I check if I’m affected by this?

Install "smartmontools" with the "brew" package manager, and run "smartctl -a /dev/disk0".

Installation is discussed in the early pages of the thread, but it's a simple process. You need to install "brew" first.

 

[TheSynchronizer]

I used Onyx to clear everything and mine seems to be writing a lot less now, I was not writing a lot anyway but I was averaging 20-30GB a day with barely very minimal usage, used the laptop for the past 6 hours and written about half that so far.

I’ve tried Onyx on my M1 mac too, and after having generally fixed my SSD writes by reducing them to a reasonable 45GB per day, Onyx has further reduced them and now i’m getting around 1GB written per hour!! That is amazingly low and I wonder if this is just temporary or if something really got fixed.

My swap is ranging from 500MB - 1GB as it was before Onyx but my writes are amazingly low. So far, I am (very positively) shocked by this

 

[Fomalhaut]

Just about everything I have read on this issue - a lot - leaves me with more questions, never answers! If we accept that all components, including SSDs, can fail, what are we left to be concerned about? A bunch of reports from people concerned about what Smart Monitor is telling them; these are not to be dismissed because most of those people appear to be well-informed and very far from stupid. But if I look at what Smart Monitor reports on my two in-service Macs I am immediately struck my some figures that are very obviously wrong. Because I am sat at my iMac just now I show its Smart Monitor report produced in the last few minutes.
Power On Hours 491? Rubbish, this machine is switched on eight hours per day seven days per week, and has been since 01/06/2019.
Power Cycles 4538? No siree.
Unsafe Shutdowns 37? Get outta here. I use a UPS precisely to prevent unsafe shutdowns.

So why should I believe anything Smart Monitor tells me? This is on an Intel Machine and it's producing gibberish, how much less is it to be trusted on Apple Silicon? As it happens, the most believable figure is what it gives for TBW, but that doesn't concern me.

I certainly believe M1 machines write a lot of data to SSD, probably more than their Intel predecessors, but does that mean they're doomed to early (within 10 years) failure? I rather think my M1 will be good for that long haul, and that you can buy with a reasonable level of confidence. After all, if there's a software problem than Apple will fix it (obviously without admitting there's an issue!) and if there's a hardware problem (affecting a small number or users...) Apple will similarly offer a fix.

Don't worry, be happy, give Apple your money.

View attachment 1743521

I think the numbers may depend on the power cycle of the SSD, not the whole computer.

"power on hours" may be lower that actual machine power-on hours if you have "put hard disks to sleep when possible" enabled in your Energy Saver settings.

similarly, the SSD may power on and off many times while the machine is turned on, which would greatly increase the number of power cycles.

"unsafe shutdown" may refer to some internal test of the SSD - e.g powering off just as a write is about to start...

 

[pcharlesmorphy]

Microsoft Word and Excel have already been optimized, right?

Excel I don't know, but Word doesn't seem to.
In any case, both applications have their version for Apple Silicon.

iMovie doesn't seem to be working well either.

It's not clear to me that Big Sur is to blame, it may be something with the applications, but I dare say that somehow, Big Sur is not optimizing memory management correctly.

What happened to me today confirms this.

 

[Donga120]

I’ve tried Onyx on my M1 mac too, and after having generally fixed my SSD writes by reducing them to a reasonable 45GB per day, Onyx has further reduced them and now i’m getting around 1GB written per hour!! That is amazingly low and I wonder if this is just temporary or if something really got fixed.

My swap is ranging from 500MB - 1GB as it was before Onyx but my writes are amazingly low. So far, I am (very positively) shocked by this

Mine has been on standby for over 12 hours and has used less than 4GB, was never like this before Onyx so something has changed.

 

[Jack Neill]

I just used Sensei to check the numbers on my main Macs:

Hackintosh with 9 month old 1TB Sabrent Rocket Q I have is at 5.27TB written as of today. 10.15.7

My M1/8/512 Air was manufactured on 11.02.20 and has 1.55TB written as of today. I've had it about 2 months and it was Open Box when I bought. Had 2 battery cycles. 11.2.3 I will assume 3 months of use.

My 2020 i5/8/512 Air was manufactured on 1.26.20 but purchased in May of 2020, also Open Box. Had 4 battery cycles on it. 4.5TB written as of today. 10.15x-11.2.3. I will assume 10 months of real use.

Rocket Q=.59TB a month
M1= .52TB a month
i5=.45 TB a month

I would say that the wear on MY M1 is on track to be normal. I am a regular user of Onyx on all machines FWIW.

YMMV

I still say let it melt and I'll buy a new one when it does.

 

[vcsyc]

Mine has been on standby for over 12 hours and has used less than 4GB, was never like this before Onyx so something has changed.

Interesting. What did you do with Onyx?

 

[Never mind]

Microsoft Word and Excel have already been optimized, right?

Yes....

Is Apple Silicon ready?

The Complete Guide to MacOS Apps Optimized for Apple Silicon Macs and the M1/M2 Processors

isapplesiliconready.com

 

[Silvestru Hosszu]

I followed closely this thread and tried to see who/what the culprit is.
I have a MBP 16 i9/16/1TB and one M1 MBP 16/512. Both on the latest BS.
I monitored them with activity monitor which being a first party app should be dependable.
To my surprise, the M1 write much LESS on the ssd (same software was used) then the intel MBP.
The big difference is related to sleep.
The M1 wrote around 4gb in 24h (kernel_task being the main culprit) and the intel 16 MBP wrote around 35GB in the same 24h.
The big difference is that 31gb from the 35 total were wrote during SLEEP.
It took about 5-10 secs for the intel MBP to wake, so I think that there is some kind of hibernation going on here.
My preliminary conclusion is that the problem is BigSur related not M1.

 

[CMMChris]

Another update from my M1 MBP. Still all fine, no strange behavior. In use since November.

 

[Spudlicious]

I think the numbers may depend on the power cycle of the SSD, not the whole computer.

"power on hours" may be lower that actual machine power-on hours if you have "put hard disks to sleep when possible" enabled in your Energy Saver settings.

similarly, the SSD may power on and off many times while the machine is turned on, which would greatly increase the number of power cycles.

"unsafe shutdown" may refer to some internal test of the SSD - e.g powering off just as a write is about to start...

Fair points. I wonder if there would be any point putting an SSD, as opposed to a motor-driven HD, to sleep. But I don't know and the utility doesn't tell me what any of the terms actually mean.The definitive skinny on this matter can only come from Apple, and they're keeping a sphinx-like silence. As they do . My instinct to be confident in my M1 perhaps comes from my time as a TV repairman in the seventies and eighties, if a TV's picture and sound were OK then it was OK, faults made themselves obvious with symptoms and sometimes with smoke! Similarly, my M1 Air is working absolutely fine, so........

 

[Donga120]

Interesting. What did you do with Onyx?

My usage was very low to begin with, averaging under 1TB a month. However since running Onyx I've averaged about 15GB in 2 days with higher usage, albeit still light but longer hours.

May be worth others trying to run Onyx and see hoow the results turn out.

Go to Maintenance tab > check everything other than:

- Spotlight Index (worth noting I don't use this and have it turned off in Settings)
- Mailboxes in Mail (again don't have lots of emails)
- Disk Positions on Desktop

For the Internet options, all are checked apart from:

- Bookmark icons
- Form values
- Instant messaging conversations

 

[TheSynchronizer]

The M1 wrote around 4gb in 24h (kernel_task being the main culprit) and the intel 16 MBP wrote around 35GB in the same 24h.
The big difference is that 31gb from the 35 total were wrote during SLEEP.

Interestingly, as much as I've followed all the tips posted throughout here to significantly reduce my M1 MBP SSD writes from 6-12GB to between 1-2GB per hour during use, my M1 Mac writes noticeably more to the SSD during sleep, than it does when the system is running.

Right now, I'm at 15 hours uptime, and 34GB written to the SSD. This works out to just over 2GB per hour. The last time I checked I was at 12 hours uptime, with 18GB written. In 3 hours of sleep (on A.C.), it wrote a whole additional 16GB!

So my M1 mac wrote 18GB during my first 12 hours of uptime. This consisted of charging the macbook, then using it on battery, then sleeping it unplugged, then using it some more on A.C., basically my regular usage. Then right before sleeping it while plugged in I checked the writes at 12 hours of reported uptime.

This leads me to believe there is some sort of bugged write loop in Big Sur A.C. sleep - whether its consistent or occasional is unclear. This not happening while the mac was unplugged also points that the issue lies in sleep when the macbook is plugged in.

Even though currently i'm at write rates that I'm perfectly comfortable with, I'd still rather not have it write unnecessarily during plugged in sleep.

 

[IceStormNG]

Interestingly, as much as I've followed all the tips posted throughout here to significantly reduce my M1 MBP SSD writes from 6-12GB to between 1-2GB per hour during use, my M1 Mac writes noticeably more to the SSD during sleep, than it does when the system is running.

Right now, I'm at 15 hours uptime, and 34GB written to the SSD. This works out to just over 2GB per hour. The last time I checked I was at 12 hours uptime, with 18GB written. In 3 hours of sleep (on A.C.), it wrote a whole additional 16GB!

macOS has to write the whole RAM to disk when it hibernates. If you have 16GB of RAM, that's 16GB it has to write.

 

[TheSynchronizer]

macOS has to write the whole RAM to disk when it hibernates. If you have 16GB of RAM, that's 16GB it has to write.

Okay that does make sense and accounts for atleast some of those writes, although I do have 8GB of ram and not 16.

As long as my writesGB / uptime < 3GB its a non issue anyway, as I expect the SSD to survive atleast 20 years at those rates, but its definitely an observation ill look into to see if it would spiral out of control if I did something else.

 

[Mike Boreham]

Not sure about the sleep theory. My M1 MBA has been on 24/7 since Nov 21st, and has never been asleep. It has 13.4 TBW. Not as bad as some but higher than many.

 

[TheSynchronizer]

Not sure about the sleep theory. My M1 MBA has been on 24/7 since Nov 21st, and has never been asleep. It has 13.4 TBW. Not as bad as some but higher than many.

I own my M1 MBP since launch too and I have 21.2 TBW (with sleeps). But you're right, it is just a theory and honestly I doubt it affects everyone even if there is an issue during sleep.

 

[Maximara]

Just about everything I have read on this issue - a lot - leaves me with more questions, never answers! If we accept that all components, including SSDs, can fail, what are we left to be concerned about? A bunch of reports from people concerned about what Smart Monitor is telling them; these are not to be dismissed because most of those people appear to be well-informed and very far from stupid. But if I look at what Smart Monitor reports on my two in-service Macs I am immediately struck my some figures that are very obviously wrong. Because I am sat at my iMac just now I show its Smart Monitor report produced in the last few minutes.
Power On Hours 491? Rubbish, this machine is switched on eight hours per day seven days per week, and has been since 01/06/2019.
Power Cycles 4538? No siree.
Unsafe Shutdowns 37? Get outta here. I use a UPS precisely to prevent unsafe shutdowns.

So why should I believe anything Smart Monitor tells me? This is on an Intel Machine and it's producing gibberish, how much less is it to be trusted on Apple Silicon? As it happens, the most believable figure is what it gives for TBW, but that doesn't concern me.

I certainly believe M1 machines write a lot of data to SSD, probably more than their Intel predecessors, but does that mean they're doomed to early (within 10 years) failure? I rather think my M1 will be good for that long haul, and that you can buy with a reasonable level of confidence. After all, if there's a software problem than Apple will fix it (obviously without admitting there's an issue!) and if there's a hardware problem (affecting a small number or users...) Apple will similarly offer a fix.

Don't worry, be happy, give Apple your money.

View attachment 1743521

Ok with 1% in 2 months (keep it simple) we have 100% at 200 months or 16.6 years. With a 14.5 TB at 1% we get a drive TBW of 1450...a little high but not totally insane like some numbers we have seen with the back check.

 

[Maximara]

Not sure about the sleep theory. My M1 MBA has been on 24/7 since Nov 21st, and has never been asleep. It has 13.4 TBW. Not as bad as some but higher than many.

November 21 to March 14 is 113 days or ~0.119 TB/day. The SSD Endurance Experiment: They’re all dead puts this in prospective. Using the worst number (200TB) gets us 1680 days or 4.6 years. More than enough time to get an external SSD, boot from that, and continue on your merry way.

 

[wirtandi]

Now we have people reporting it may be Big Sur related, not M1, because their Intel based machines are affected.

Oh gosh, I surely hope Apple is silently investigating this. It is unfair that us customers have to do all the investigations when they are the one who should be informing us of the issue

 

[Mike Boreham]

November 21 to March 14 is 113 days or ~0.119 TB/day. The SSD Endurance Experiment: They’re all dead puts this in prospective. Using the worst number (200TB) gets us 1680 days or 4.6 years. More than enough time to get an external SSD, boot from that, and continue on your merry way.

Thanks. I am more curious than worried about this issue. I am in the camp that thinks there will either be an explanation or a software fix, or a repair program, with one of the first two the most likely.