ssd swap - high usage of Terabytes Written

[Maximara]

A more useful video IMHO is Is There A Problem? Swap Memory on M1 Macs which treats Apple's SSD TBW as a black box. The conclusion from various posts on the threads the author read are:

256 GB: 1400 TBW
512 GB: 2000 TBW
1 TB: 2500 TBW
2 TB: 3000 TBW

The formula if you want to do this calculation yourself is TB*100/percentage = TBW and here is how it is derived:

Base formula: TB/TBWx100 = percentage.
TB = percentage/100 * TBW
TB/(percentage/100) = TBW
TB*100/percentage = TBW

So if 3.5% has been used for 5.32 then the TBW is 152 which giving the rounding is 150. If the percentage is 0 this equation will fail so kick it up to 0.4999% for the worst case situation. This would make macrumors member's 4.49 TB translate into 898.2 TBW at worst. But it could be 0.000...1% so until we get an actual percentage over 0% (0.5 or higher) we're guessing.

 

[Forti]

I saw about 20GB written while my machine was supposedly sleeping...powernap is disabled, so I'm guessing it's local TimeMachine backup (to the SSD).

I have disabled TimeMachine backup form the very beginning. There must be something else.

 

[Forti]

A more useful video IMHO is Is There A Problem? Swap Memory on M1 Macs which treats Apple's SSD TBW as a black box. The conclusion from various posts on the threads the author read are:

256 GB: 1400 TBW
512 GB: 2000 TBW
1 TB: 2500 TBW
2 TB: 3000 TBW

The formula if you want to do this calculation yourself is TB/(100*percentage) = TBW

So if 3.5% has been used for 5.32 then the TBW is 152 which giving the rounding is 150. If the percentage is 0 this equation will fail so kick it up to 1%.

There are already some dead ssd on twitter after ~~700 TBW per 256gb.

 

[inversed]

Outside of Lightroom Classic i'm totally unaffected. My Mac Mini can sleep, i can work, etc.. I never see large amount of disk write that is not correlated to my usage.
My Mac Mini is a fresh install, without iCloud and TimeMachine running. The only non work related app is Little Snitch (latest version).

 

[Maximara]

There are already some dead ssd on twitter after ~~700 TBW per 256gb.

Do these posts show the last percentage used before the drive died? I ask because is could have been the controller not the SSD itself that went walkabout. Also what as their workflow like? Were they, as one video suggests, trying to do work that really required 16 GB on a 8 GB Mac and have the OS page the SSD into oblivion?

 

[jdb8167]

Do these posts show the last percentage used before the drive died? I ask because is could have been the controller not the SSD itself that went walkabout.

The controller is on the M1 SoC but the drive could have died before it reached the warrantied TBW. I'd definitely be curious what the percentage used was because it would tell us what the nominal TBW rating is.

 

[chouseworth]

I've got a question for the more knowledgeable folks on this forum re the expected life of the internal SSD in TBW. Say you have a 256G SSD that you have stuffed with 225G of your own data, apps, etc. Then say there is another guy with the same 256G SSD and same memory swap activity who has his only stuffed with 100G of data, apps, etc. Does his SSD have a longer expected life in TBW because more of his drive is "empty" and available for memory swap? Or does the OS dynamically take all the existing data and shift it around in order that the space available for swap essentially remains the same between the two SSDs, essentially giving each the same expected life in TBW? It seems that with a drive that is near full, and the dynamic shifting described above, all of the shifting around of existing data, apps, etc, could contribute to a lower expected life even with relatively low memory swap activity. But that is just my layman's view.

 

[Maximara]

The controller is on the M1 SoC but the drive could have died before it reached the warrantied TBW. I'd definitely be curious what the percentage used was because it would tell us what the nominal TBW rating is.

Also as with any drive there is the chance the one had a defect which since it hasn't been a year should be replaced by Apple as a years has not passed on the default warranty. FTR people have reported Windows SSDs going nuts too though that was partly because Windows was doing something incredibly stupid - defragging the SSD. What programmer with any sense allows the OS to defrag an SSD?!

 

[Dockland]

Do these posts show the last percentage used before the drive died? I ask because is could have been the controller not the SSD itself that went walkabout. Also what as their workflow like? Were they, as one video suggests, trying to do work that really required 16 GB on a 8 GB Mac and have the OS page the SSD into oblivion?

How many of these "dead" SSD are known? Are there only one single case?

 

[Maximara]

I've got a question for the more knowledgeable folks on this forum re the expected life of the internal SSD in TBW. Say you have a 256G SSD that you have stuffed with 225G of your own data, apps, etc. Then say there is another guy with the same 256G SSD who has his only stuffed with 100G of data, apps, etc. Does his SSD have a longer expected life in TBW because more of his drive is "empty" and available for swap? Or does the OS dynamically take all the existing data and shift it around in order that the space available for swap essentially remains the same between the two SSDs, essentially giving each the same expected life in TBW? It seems that with a drive that is near full, all of the shifting around of existing data, apps, etc, could contribute to a lower the expected life. But that is just my layman's view.

Your layman's view seems to be spot on as the larger the SSD the larger the TBW so logically the smaller the amount used the more can be used to replace the cells that die.

 

[Maximara]

How many of these "dead" SSD are known? Are there only one single case?

There is the one that appeared in Tidbits but it was for an Intel Mac and assuming it was gotten new the SSD lasted around 4 to 5 years. Also there is a lot of data we don't have for that example.

 

[k-hawinkler]

I've got a question for the more knowledgeable folks on this forum re the expected life of the internal SSD in TBW. Say you have a 256G SSD that you have stuffed with 225G of your own data, apps, etc. Then say there is another guy with the same 256G SSD and same memory swap activity who has his only stuffed with 100G of data, apps, etc. Does his SSD have a longer expected life in TBW because more of his drive is "empty" and available for memory swap? Or does the OS dynamically take all the existing data and shift it around in order that the space available for swap essentially remains the same between the two SSDs, essentially giving each the same expected life in TBW? It seems that with a drive that is near full, and the dynamic shifting described above, all of the shifting around of existing data, apps, etc, could contribute to a lower expected life even with relatively low memory swap activity. But that is just my layman's view.

Great analysis.

The more unused storage locations there are on an SSD the better for its longevity.
Trditionally the recommendation has been to not use more than about 75%, give or take a few.
That’s one of the reasons why I chose a 2TB SSD for myM1 Mac mini.
Another reason is it can’t be upgraded later, except for a so called “forklift“ upgrade.
That is you throw the entire computer out and get a new one.
Not an attractive solution for me.

 

[qap]

Remember that on ARM Macs is not 1x 256/512...GB drive but are 2x 128/256/512...GB drives

 

[k-hawinkler]

Remember that on ARM Macs is not 1x 256/512...GB drive but are 2x 128/256/512...GB drives

Thanks.
Do you have any idea what algorithm is used to spread the data across available storage locations?

 

[qap]

Thanks.
Do you have any idea what algorithm is used to spread the data across available storage locations?

No... this is the point, I've already written, this could be the cause for some weird readings from smartctl. But maybe is not and the numbers are correct. At the moment I have no idea. It's hard to investigate since Apple is using proprietary controller for RAM and SSD.

 

[Fishrrman]

I don't have an m-series Mac yet.
I do have a 2018 Intel Mini with 16gb RAM, running Mojave.

What I'd like to know is:
Can you TURN OFF VM disk swapping on the m-series Macs?

I've turned it off on my Mini using terminal:
sudo launchctl unload -wF /System/Library/LaunchDaemons/com.apple.dynamic_pager.plist
(there may be other commands that do the same)

Then, reboot.
After doing this, I use this command:
sysctl vm.swapusage

And terminal gives this result:
vm.swapusage: total = 0.00M used = 0.00M free = 0.00M

If I go to Activity monitor, it shows:
Swap used: 0 bytes.

After doing this, one must be careful not to "overload" the RAM, but I've had no problems and I get NO CRASHES from running out of memory.

Since there seem to be numerous "experimenters" in this thread, I was wondering if one or more could try TURNING OFF VM DISK SWAPPING (again, if that's possible in Big Sur/m1) and see how that impacts the disk loading...

 

[Baff]

A more useful video IMHO is Is There A Problem? Swap Memory on M1 Macs which treats Apple's SSD TBW as a black box. The conclusion from various posts on the threads the author read are:

256 GB: 1400 TBW
512 GB: 2000 TBW
1 TB: 2500 TBW
2 TB: 3000 TBW

The formula if you want to do this calculation yourself is TB*100/percentage = TBW and here is how it is derived:

Base formula: TB/TBWx100 = percentage.
TB = percentage/100 * TBW
TB/(percentage/100) = TBW
TB*100/percentage = TBW

So if 3.5% has been used for 5.32 then the TBW is 152 which giving the rounding is 150. If the percentage is 0 this equation will fail so kick it up to 0.4999% for the worst case situation. This would make macrumors member's 4.49 TB translate into 898.2 TBW at worst. But it could be 0.000...1% so until we get an actual percentage over 0% (0.5 or higher) we're guessing.

Sorry, but the calculation isn't as simple as that. Or, the number given by both DriveDX and Smartmontools is incorrect. At 2TB written, I was at 2%. Now I am at 3.7TB written and still at 2%.

 

[Maximara]

Great analysis.

The more unused storage locations there are on an SSD the better for its longevity.
Trditionally the recommendation has been to not use more than about 75%, give or take a few.
That’s one of the reasons why I chose a 2TB SSD for myM1 Mac mini.
Another reason is it can’t be upgraded later, except for a so called “forklift“ upgrade.
That is you throw the entire computer out and get a new one.
Not an attractive solution for me.

This seems to forget that booting from an external SSD is an option. The inability to upgrade RAM is an issue to everything but the Pro and the old Intel MacMinis.

 

[k-hawinkler]

This seems to forget that booting from an external SSD is an option. The inability to upgrade RAM is an issue to everything but the Pro and the old Intel MacMinis.

No, I didn’t forget. Mine is set up to do just that.
But it is less convenient to operate because of heightened macOS security in Big Sur.

 

[eshroom]

Seems turning spotlight off had 0 effect for me. In 3 power on hours it's written 600gb - a faster rate than when I had spotlight turned on...

The reason this issue concerns me is a) I have the 8gb model and b) I am the kind of person that looks after my laptops and expect them to keep running for years. My mum still happily uses my 2012 Macbook Air daily with 0 issues, even after a coffee spill a few years ago.

View attachment 1737743

View attachment 1737742

Dramatic fall in SSD usage achieved simply by cutting the number of tabs open on Chrome. I am terrible for having a dozen plus tabs open at any one time on Chrome, and I know from experience that Chrome is terrible when it comes to optimising memory and CPU usage with multiple tabs open.

I have cut this down to 5 tabs open at any one time, and while this is very annoying as I don't like having to moderate the way I use my mac, it has solved the problem for me.

Previous SSD writes: 0.08TB per hour
Now: 0.022TB per hour

What all this shows is that cutting spotlight or any other workaround just won't work. The only way to reduce SSD writes is to make sure you are optimising the memory use of your machine. I chose 8gb memory for the first time in nearly a decade specifically because of the SSDs great swap performance, clearly if you want your machine to last a long time and regularly perform memory intensive tasks, you need to upgrade your memory when buying.

 

[Maximara]

Sorry, but the calculation isn't as simple as that. Or, the number given by both DriveDX and Smartmontools is incorrect. At 2TB written, I was at 2%. Now I am at 3.7TB written and still at 2%.

Remember "2%" could be anything from 1.5% to 2.49......9%. Using the implied TB*100/percentage = TBW results in this:

2*100/1.5 = ~133 TBW; 2*100/2.49...9 = ~80 TBW
3.7*100/1.5 = ~246.6 TBW; 2*100/2.49...9 = ~148 TBW

Since there is no overlap between the ranges something has to be wonked - either in the formula or in the data being given by the tools.

More over Eshroom's 10.6 with 0% works out to be, at worst 2124 TBW (10.6*100/0.499). Without the size fo the drives that is all I can do with the numbers but clearly something is way out of wack. There just NOT should be that much of a variance in the black box TBW.

 

[Baff]

Remember "2%" could be anything from 1.5% to 2.49......9%. Using the implied TB*100/percentage = TBW results in this:

2*100/1.5 = ~133 TBW; 2*100/2.49...9 = ~80 TBW
3.7*100/1.5 = ~246.6 TBW; 2*100/2.49...9 = ~148 TBW

Since there is no overlap between the ranges something has to be wonked - either in the formula or in the data being given by the tools.

More over Eshroom's 10.6 with 0% works out to be, at worst 2124 TBW (10.6*100/0.499). Without the size fo the drives that is all I can do with the numbers but clearly something is way out of wack. There just NOT should be that much of a variance in the black box TBW.

I do have a theory on why I am at 2% with such low TB written. I bought this Mini as a refurb last week. If as part of the refurb process, Apple writes zeros to the whole drive (so that new owners can't recover data from the previous owner), that might well reduce the life of the drive by 1 or 2%. So, I don't think the equation for Lifetime Left/Used is as simple as they describe it in that video.

 

[tab0reqq]

My DriveDx Trial has ended, but again... up to some point there was 100GB/night (unplugged, sleep). Then today morning ~900GB written by the kernel task. I just can't see the pattern ;/

More or less for me its:
- small writes
- small writes
- it goes all-in and makes almost 1TB writes in a day.

Plugged or unplugged. With or without an external monitor. That's just plain stupid and I think that Apple should step in and say:
"we know it, it's ok" or "we investigate" or "here is a fix".

 

[Jinbei]

new update incoming, maybe a silent fix on the way

 

[Dockland]

new update incoming, maybe a silent fix on the way

Yes, just got 11.2.3 Downloading it now.

3,37 GB in size