ssd swap - high usage of Terabytes Written

[k-hawinkler]

See my previous post about Class Action. The responsible manufacturer, in this case, is Apple (regardless of who made the drive). Apple CHOSE to define a "percent life" value via the SMART data in the drive. If the life of the drive were found to be drastically lower for a large percentage of the purchasers, an implied warranty would be asserted. There is next to a zero chance that the value (which if anything is stated as being a conservative estimate of lifespan) is substantially lower than the way it is coded in SMART AND that Apple is ignoring a real issue that has been widely reported.

Great.

 

[ambient_light]

Have you made any of the adjustments suggested in this thread?

If you refer to various suggestions about how to limit the effect on OS write activity from applications (browsers) - it‘s a workaround, which may temporarily help to hide problem for some usage scenarios, but not solve it. Think about applications as clients of OS services, it’s not their responsibility to optimize memory management algorithms, specifically swapping, they are responsible only for their own memory usage, and cannot control excessive swap pages writes OS is doing.

So, the actual bug is within OS kernel, as was demonstrated few times here, and when Apple is going to fix it is anyone’s guess.

 

[Mike Boreham]

Based on his response, it seems that "Data Written" in Activity Monitor gives the total of ALL connected drives UNTIL the drive is disconnected, then falls back to the number for the host drive only. In any case, the "real" number is easily extracted by tools or terminal commannds.

I have done some testing on how AM treats externals and confirmed the above and can add a bit.

1. Starting point was my M1 MBA with a 2TB external connected as usual. AM showed 206GB written in last 43h uptime.
2. Disconnected 2TB. AM went down to 180GB, down 25GB.
3. Reconnected 2TB. AM did not go back to 206GB. stayed at 180GB.
4. Connected a second external and copied 103GB to it from the 2TB drive. AM went up by 98GB. (why not 103GB?)
5. Disconnected the second external. AM went back to 181GB.
6. Disconnected 2TB. AM stayed at 181GB.
7. Reconnected second external. AM stayed at 181GB.
8. Reconnected 2TB. AM stayed at 181GB.

Conclusions:
-Writes to externals are shown by AM, but the write history is not stored so it doesn't come back on reconnection.
-With no externals connected AM shows just the writes to the internal drive.

In my case, for the last 43hours, my writes to the internal have been 181GB, and an additional 25GB to my normally connected 2TB drive. 4.3GB/h seems to be regarded as OK.

I asked the question about AM in relation to smartmontools and DriveDX because it seemed AM might be bit easier to have a quick daily look at AM for what was happening in the short term, compared to smartmon and DriveDx (which only report TBW to one decimal place, so OK for assessing over a long period but not a short one, without more arithmetic from the units numbers).

 

[k-hawinkler]

I have done some testing on how AM treats externals and confirmed the above and can add a bit.

1. Starting point was my M1 MBA with a 2TB external connected as usual. AM showed 206GB written in last 43h uptime.
2. Disconnected 2TB. AM went down to 180GB, down 25GB.
3. Reconnected 2TB. AM did not go back to 206GB. stayed at 180GB.
4. Connected a second external and copied 103GB to it from the 2TB drive. AM went up by 98GB. (why not 103GB?)
5. Disconnected the second external. AM went back to 181GB.
6. Disconnected 2TB. AM stayed at 181GB.
7. Reconnected second external. AM stayed at 181GB.
8. Reconnected 2TB. AM stayed at 181GB.

Conclusions:
-Writes to externals are shown by AM, but the write history is not stored so it doesn't come back on reconnection.
-With no externals connected AM shows just the writes to the internal drive.

In my case, for the last 42hours, my writes to the internal have been 181GB, and an additional 25GB to my normally connected 2TB drive. 4.3GB/h seems to be regarded as OK.

I asked the question about AM in relation to smartmontools and DriveDX because it seemed AM might be bit easier to have a quick daily look at AM for what was happening in the short term, compared to smartmon and DriveDx (which only report TBW to one decimal place, so OK for assessing over a long period but not a short one, without more arithmetic from the units numbers).

Many thanks for a very interesting report.
I simply unmounted but did not disconnect external drives from my M1 Mac mini and did not see the Data Written value in the Activity Monitor going down.
So next time I will actually disconnect the external drives and see what happens then with my setup.
Thanks again indeed for your experience and your report. 👍

 

[Mike Boreham]

Many thanks for a very interesting report.
I simply unmounted but did not disconnect external drives from my M1 Mac mini and did not see the Data Written value in the Activity Monitor going down.
So next time I will actually disconnect the external drives and see what happens then with my setup.
Thanks again indeed for your experience and your report. 👍

Thanks for the feedback.
Yes I actually physically disconnected the externals. I didn't look at what AM showed after just unmounting, and will do next time.

 

[Maximara]

Apple's lack of response AFTER the problem has been reported would make them liable for a giant class-action suit.

It might make them the target of such a suit but given the complexity of computer software and the fact there are people not having the problem it would be hard for Apple to loose the case. And even if Apple did some how loose the appeals court would likely reduce any award, kick it back to the lower court for a redo, or throw the thing out entirely.

Time Limits for Filing Product Liability Cases: State-by-State shows that such a lawsuit would be a complicated mess as the states all have wildly different statute of limitations.

Lawsuits, especially class action lawsuits are not the magic bullet a lot of people think they are.

 

[Maximara]

See my previous post about Class Action. The responsible manufacturer, in this case, is Apple (regardless of who made the drive). Apple CHOSE to define a "percent life" value via the SMART data in the drive. If the life of the drive were found to be drastically lower for a large percentage of the purchasers, an implied warranty would be asserted. There is next to a zero chance that the value (which if anything is stated as being a conservative estimate of lifespan) is substantially lower than the way it is coded in SMART AND that Apple is ignoring a real issue that has been widely reported.

Accuse the following. Apple really likes using ALL CAPS for its warranties.

"TO THE EXTENT PERMITTED BY LAW, THIS WARRANTY AND THE REMEDIES SET FORTH ARE EXCLUSIVE AND IN LIEU OF ALL OTHER WARRANTIES, REMEDIES AND CONDITIONS, WHETHER ORAL, WRITTEN, STATUTORY, EXPRESS OR IMPLIED. APPLE DISCLAIMS ALL STATUTORY AND IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE AND WARRANTIES AGAINST HIDDEN OR LATENT DEFECTS, TO THE EXTENT PERMITTED BY LAW."

The state laws are, to put it bluntly, a mess with regards to the very acceptance of an implied warranty and may override part of that.

More over "Unfortunately, regardless of the storage technology, there is no industry-wide standard to tell you which numbered SMART attribute describes a specific physical property of a drive The attribute descriptors will vary between SSD and HDD, and even between the various SSD vendors.

There are several third-party utilities that can retrieve and report a drive’s SMART data, often available in freeware and shareware. However, unless the third-party software vendor consults with the drive vendor on correct SMART attributes, their definitions and thresholds (when applicable) are likely to mislabel the attributes and can lead to false-positive or false-negative failure reports. (...) Incorrectly reported or misinterpreted SMART data can lead to incorrect conclusions which, can, unfortunately, lead to the return of a perfectly functional drive." - SSDs and SMART Data

 

[leons]

I won't argue the legal case here and will stipulate your detailed interpretation of warranty law and third-party tool statements. As I said, I use no third-party tools to interpret the SMART data and only go by the raw data as reported by Apple's own OS. Additionally, I was not prejudging a potential court case result, but simply stating my informed opinion that Apple would never allow the PR nightmare to even allow such a suit to be initiated. I rest my case. I will not discuss legal issues further in this thread (and I apologize for my part in bringing them up without realizing the rabbit-hole I was creating). I suggest we bring this thread back to its original purpose: to help people who have seemingly excessive writes to their SSD to reduce them until Apple and/or third-party vendors make adjust their software.

 

[leons]

I have done some testing on how AM treats externals and confirmed the above and can add a bit.

1. Starting point was my M1 MBA with a 2TB external connected as usual. AM showed 206GB written in last 43h uptime.
2. Disconnected 2TB. AM went down to 180GB, down 25GB.
3. Reconnected 2TB. AM did not go back to 206GB. stayed at 180GB.
4. Connected a second external and copied 103GB to it from the 2TB drive. AM went up by 98GB. (why not 103GB?)
5. Disconnected the second external. AM went back to 181GB.
6. Disconnected 2TB. AM stayed at 181GB.
7. Reconnected second external. AM stayed at 181GB.
8. Reconnected 2TB. AM stayed at 181GB.

Conclusions:
-Writes to externals are shown by AM, but the write history is not stored so it doesn't come back on reconnection.
-With no externals connected AM shows just the writes to the internal drive.

In my case, for the last 43hours, my writes to the internal have been 181GB, and an additional 25GB to my normally connected 2TB drive. 4.3GB/h seems to be regarded as OK.

I asked the question about AM in relation to smartmontools and DriveDX because it seemed AM might be bit easier to have a quick daily look at AM for what was happening in the short term, compared to smartmon and DriveDx (which only report TBW to one decimal place, so OK for assessing over a long period but not a short one, without more arithmetic from the units numbers).

Very good and VERY useful info, thanks VERY MUCH for your detailed reporting!

 

[kubilaydem]

2.85 gb write, 9 hours 25 min uptime.

Web browsing + 1 hour steam gaming.

 

[Fishrrman]

Baff wrote:
"I switched from Safari to Firefox and turned off the memory/cache. This was disastrous for me and my 70+ tabs..."

I gotta ask:
What does one need 70+ tabs open for...?
For that matter, why would one need 10 tabs...?

<<< The old Fishrrman doesn't use tabs AT ALL. I'm morally opposed to them...

 

[k-hawinkler]

Baff wrote:
"I switched from Safari to Firefox and turned off the memory/cache. This was disastrous for me and my 70+ tabs..."

I gotta ask:
What does one need 70+ tabs open for...?
For that matter, why would one need 10 tabs...?

<<< The old Fishrrman doesn't use tabs AT ALL. I'm morally opposed to them...

Well someone else explained to me to keep references alive for their research. I wish I had that option 50 years ago when I got into research.

Besides the M1 computer should easily support that kind of activity.Don’t you think so?

 

[gank41]

Those of you making any changes in Terminal or changing permissions on folders, just make sure you make a note of those changes so you can set things back to normal once Apple fixes the issue.

 

[gank41]

Baff wrote:
"I switched from Safari to Firefox and turned off the memory/cache. This was disastrous for me and my 70+ tabs..."

I gotta ask:
What does one need 70+ tabs open for...?
For that matter, why would one need 10 tabs...?

<<< The old Fishrrman doesn't use tabs AT ALL. I'm morally opposed to them...

That’s why I use Safari, so I can add things to my Reading List and then keep a minimal number of tabs open.

 

[Maximara]

I won't argue the legal case here and will stipulate your detailed interpretation of warranty law and third-party tool statements. As I said, I use no third-party tools to interpret the SMART data and only go by the raw data as reported by Apple's own OS.

As stated there is "no industry-wide standard to tell you which numbered SMART attribute describes a specific physical property of a drive The attribute descriptors will vary between SSD and HDD, and even between the various SSD vendors." So you have a black box of the SSD itself (whose brand we don't know) and the controller Apple uses. The only thing we have to work with is Activity Monitor and it seems to be given weird numbers.

mds_stores (part of Spotlight) writes 415.75 GB but Data Written shows only 178.15 GB written or 42.850% and there are other processes writing data. The numbers just don't make sense as I don't think there is any kind of compression going on before the data is actually written to the SSD.

 

[leons]

To be clear to those who may be new to this thread: for the vast majority of users and usage cases, making adjustments per the information in this thread will reduce SSD writes to an amount that (barring catastrophic failure) assures that the SSD will last well past the useful life of the machine, by any measure. Many who had the issue have reported (here and elsewhere) success with these adjustments. Of course, I and others, await a permanent fix with no adjustments needed. For the few that use specific disk-intensive applications (Lightroom, etc.), many have workarounds posted in the application-specific forum (e.g. Adobe). It is also possible that these specific applications will cease to be an issue when they are ported to native M1.

 

[Baff]

Baff wrote:
"I switched from Safari to Firefox and turned off the memory/cache. This was disastrous for me and my 70+ tabs..."

I gotta ask:
What does one need 70+ tabs open for...?
For that matter, why would one need 10 tabs...?

<<< The old Fishrrman doesn't use tabs AT ALL. I'm morally opposed to them...

Occasionally in movies you see some computer guy with 50 monitors all over the wall showing different things. That is me, but all in tabs on 1 monitor.

Many of the tabs are static reference (data, maps, etc.), they never need to be reloaded.
Many of the tabs are temporary. I maintain a website devoted to sci-fi/fantasy tv start dates. I open a bunch of tabs related to various tv shows when I update my site. After I update, I close all those tabs.
When I shop for an item, I often open a bunch of tabs to compare specs, price, rating, etc.
I have a bunch of tabs open for a wide variety of news sites, many grouped on specific subjects (computers, tv, covid, tesla (starlink, solar), games, etc.)
Etc. etc.
I use almost every tab every day. Most tabs get used dozens or even hundreds of times a day. The tabs I don't use every day, I closed to reduce memory usage.

I can no longer even imagine using only 1 tab.

 

[leons]

Occasionally in movies you see some computer guy with 50 monitors all over the wall showing different things. That is me, but all in tabs on 1 monitor.

Many of the tabs are static reference (data, maps, etc.), they never need to be reloaded.
Many of the tabs are temporary. I maintain a website devoted to sci-fi/fantasy tv start dates. I open a bunch of tabs related to various tv shows when I update my site. After I update, I close all those tabs.
When I shop for an item, I often open a bunch of tabs to compare specs, price, rating, etc.
I have a bunch of tabs open for a wide variety of news sites, many grouped on specific subjects (computers, tv, covid, tesla (starlink, solar), games, etc.)
Etc. etc.
I use almost every tab every day. The tabs I don't use every day, I closed to reduce memory usage.

I can no longer even imagine using only 1 tab.

The "Tab Abusers" (I am one) vs the "Bookmark/Few Tab" camps will never agree. Like so many things, it is a highly personal style for which neither is right and neither will convince the other. I'm just glad that they haven't started a "Me Too" movement for tabs!

400 tabs and it's running smoothly like a champ. No problems.
https://twitter.com/i/status/1328790134548905985

 

[Fishrrman]

k-haw wrote:
"Besides the M1 computer should easily support that kind of activity.Don’t you think so?"

I would think that 70+ tabs could be one of the reasons behind an extraordinarily high amount of disk swapping...

 

[leons]

k-haw wrote:
"Besides the M1 computer should easily support that kind of activity.Don’t you think so?"

I would think that 70+ tabs could be one of the reasons behind an extraordinarily high amount of disk swapping...

It absolutely is. That's why the Tab-Abusers amongst us worked so hard to find a siolution.

 

[meropenem]

To be clear to those who may be new to this thread: for the vast majority of users and usage cases, making adjustments per the information in this thread will reduce SSD writes to an amount that (barring catastrophic failure) assures that the SSD will last well past the useful life of the machine, by any measure. Many who had the issue have reported (here and elsewhere) success with these adjustments. Of course, I and others, await a permanent fix with no adjustments needed. For the few that use specific disk-intensive applications (Lightroom, etc.), many have workarounds posted in the application-specific forum (e.g. Adobe). It is also possible that these specific applications will cease to be an issue when they are ported to native M1.

Where is there a specific solution to Lightroom thrashing the SSD? I haven’t found one

 

[k-hawinkler]

It absolutely is. That's why the Tab-Abusers amongst us worked so hard to find a siolution.

In your sentence I would replace Tab-Abusers with Tab-Creative Inventive Users.

 

[Baff]

k-haw wrote:
"Besides the M1 computer should easily support that kind of activity.Don’t you think so?"

I would think that 70+ tabs could be one of the reasons behind an extraordinarily high amount of disk swapping...

It is, but it should not be. 100 tabs was never an issue with my 6 year old mini. This is an OS/software bug that should be fixable by Apple. If not in the upcoming update, then in the next OS in the Fall. With this work-around, I can wait that long. Sadly, it doesn't work for people getting writes from other apps.

I haven't done any hardcore gaming on my M1 yet, so I don't know if that will be an issue for me or not. Luckily, I have had a bunch of PS5 games to keep me busy. At some point I will have to download a few and see what happens. I haven't seen anyone mentioning gaming as a source of massive writes yet, but I may have missed it.

 

[Maximara]

It is, but it should not be. 100 tabs was never an issue with my 6 year old mini.

Which isn't a surprise as the Mac mini (Late 2014) had Fusion Drives rather then pure SSDs. Unless those tabs are "frequently used" they will be sitting on the HD not the SSD section IIUC how the fusion drive works.

 

[Baff]

Which isn't a surprise as the Mac mini (Late 2014) had Fusion Drives rather then pure SSDs. Unless those tabs are "frequently used" they will be sitting on the HD not the SSD section IIUC how the fusion drive works.

Good point, though I didn't have massive writes to either side of the drive.