ssd swap - high usage of Terabytes Written

[Fishrrman]

Once again, I would like to propose that one, two or three m1 owners take my suggestion to DISABLE VM disk swapping and compressed memory using terminal.

Then, run that way for a day or two, recording the relevant information.

Then, come back and report your results here.

My prediction is that this will go a long way towards "curing" the excessive swapping/disk usage problems. But I don't have an m1 Mac on which to test such predictions, myself.

 

[featherlessbird]

According to the latest Tinker Tool System, my 8GB RAM / M1 MBP has used a shocking 21% of SSD Disk life since Mid December which gives a disk life of two years or less. Actual runtime 782 hours. (Drive DX concurs and shows 234TB written to disk) My old 2018 Intel Mac Mini has used about 4%. MBP Also has a bad case of runaway Kernel Tasks.

Thoughts or ideas?

That's interesting, my 8/256GB MBA has used only 16% with 259TBW.

 

[Maximara]

That's interesting, my 8/256GB MBA has used only 16% with 259TBW.

234TB giving 21% on a 8/not given but 259TBW giving 16% on a 8/256GB?

If there ever was a red flag that something is wonked with the tools that is it, as regardless of how the percentage is measuring thing you cannot have more data written resulting in a lower percentag of usage.

 

[jdb8167]

234TB giving 21% on a 8/not given but 259TBW giving 16% on a 8/256GB?

If there ever was a red flag that something is wonked with the tools that is it, as regardless of how the percentage is measuring thing you cannot have more data written resulting in a lower percentage.

Nonsense. It has been explained to you over and over, the percent remaining doesn't have to be a linear value. Both of those numbers are reasonable.

 

[Maximara]

Nonsense. It has been explained to you over and over, the percent remaining doesn't have to be a linear value. Both of those numbers are reasonable.

And I have presented The SSD Endurance Experiment: They’re all dead article as proof the equation is linear but my post didn't even use that per the "as regardless of how the percentage is measuring things" part of my comment.

It doesn't matter if the formula is linear, logarithmic, or exponential there is no way 21% = 234TBW can agree with 16% = 259TBW. It's basic math that if 21% = x and 16% = y then because 21% > 16% than x > y.

Please tell me in what world other then negative numbers is 234 > 259?!

 

[jdb8167]

And I have presented The SSD Endurance Experiment: They’re all dead article as proof the equation is linear but my post didn't even use that per the "as regardless of how the percentage is measuring things" part of my comment.

It doesn't matter if the formula is linear, logarithmic, or exponential there is no way 21% = 234TBW can agree with 16% = 259TBW. It's basic math that if 21% = x and 16% = y then because 21% > 16% than x > y.

Please tell me in what world other then negative numbers is 234 > 259?!

There is more to SSD lifespan than the total number of bytes written. It is up to the vendor to decide how to calculate Percentage Used. They don't have to just look at the number of bytes. So you are overreacting to something you don't understand. Apple is the vendor. They designed the SSD controller and wrote the SMART monitor APIs. They get to decide what the numbers mean. It is absolutely possible for both numbers to be correct since they are only approximations anyway.

 

[IceStormNG]

Most controllers also take write amplification into account. Lifetime left is usually based on NAND writes. Not host writes. Some SSDs do not expose NAND writes separately. NAND writes are actually more important than host writes.

And therefore you can get different percentages with the same host writes. There's nothing wrong with that. It's just how it works.

And: Writing SSDs linearly full, then completely erase and repeat that over again is a very different pattern than real-world use. Wear leveling will not kick in that hard (or not at all) on constant full-disk re-writes. And therefore has lower write amplification.

Wear leveling kicks in, when certain cells have not been written to for a long time, while others are already more worn out. Then the controller moves these unchanged blocks to a more worn blocks and writes new data to the fresher cells to they wear out evenly. If you run out of reserve cells and a single cell fails, the SSD goes into read only and is considered worn out. To get the best lifetime out of the NAND, the controllers moves data around so the NAND is worn out evenly. This might cause additional writes. Depending on your usage more or less.

 

[sneeks]

My Apple refurbished Mac Mini 16/256gb arrived today and the initial reading for data units written on this one are far lower than the Mini 8/256gb it replaced, that was on 3.5TB when it arrived and this is on 345GB. Guessing the old refurb must have had more use with the original buyer. I’ve adjusted most of the settings shown in the first post and will monitor the disk write usage. The only constant data write i can currently see in Activity Monitor is for cloudd which will hopefully stabilise once all data is synced.

Edit: It's still syncing data from iCloud, currently at 50gb written today. I'd much rather the data remained on iCloud rather than wasting space on my SSD. I don’t recall it saving all of this data to the Mac Mini I returned but maybe it did so overnight and I had not realised.

 

[leons]

58GB is nothing to worry about for a day under any circumstances.

 

[sneeks]

What are you running that you are using up "a good portion of" your RAM before you even open Lightroom or Photoshop? I use the open source Stats to monitor my RAM usage on my old 2013 8 BG iMac and only go north of 70% when I watch Youtube videos, play Hearthstone, or play Raid Shadow legends.

I installed Stats onto my new 16gb Mac Mini and attach some screenshots to show the RAM use, on my 8gb Mini swap was required for these same tasks when Photoshop, Lightroom or a combination of both were open.

 

[sneeks]

58GB is nothing to worry about for a day under any circumstances.

I think that will stop writing so much once it has finished syncing my 200gb iCloud data. Doesn't look like any need for concern with excess write.

 

[Maximara]

I installed Stats onto my new 16gb Mac Mini and attach some screenshots to show the RAM use, on my 8gb Mini swap was required for these same tasks when Photoshop, Lightroom or a combination of both were open.

View attachment 1771707 View attachment 1771708 View attachment 1771710 View attachment 1771706

These all show "swap" at 0 KB which would indicate any SSD writing is not due to RAM. My 2013 iMac (8 GB) shows 308 MB swap with 79% running Raid Shadow legends and having my browser open to type this.

 

[Maximara]

There is more to SSD lifespan than the total number of bytes written.

And yet that is what nearly all these posts are harping on SSD lifespan being too short are using. Either total number of bytes written is a primary determiner of SSD lifespan or it is not. You can't have it both ways.

As for this only being a Big Sur/M1 problem read SSD wear leveling count at 90% after 7 months Neither existed in 2014.

 

[sneeks]

These all show "swap" at 0 KB which would indicate any SSD writing is not due to RAM. My 2013 iMac (8 GB) shows 308 MB swap with 79% running Raid Shadow legends and having my browser open to type this.

Yes these are all from my 16gb Mini, running these on the 8gb caused swap to be used but unfortunately that’s now gone so I cannot show screenshots of that. RAM usage was the same on the 8gb so did require Swap.

 

[Maximara]

Yes these are all from my 16gb Mini, running these on the 8gb caused swap to be used but unfortunately that’s now gone so I cannot show screenshots of that. RAM usage was the same on the 8gb so did require Swap.

Ah. I missed the "Used" part and that tells us something critical. With the exception of the last picture everything is north of 8 GB so on a 8 GB swap would be happening with more RAM resulting more swap. I wonder how many of the people having problems are running programs that need RAm way beyond what they have.

 

[sneeks]

Ah. I missed the "Used" part and that tells us something critical. With the exception of the last picture everything is north of 8 GB so on a 8 GB swap would be happening with more RAM resulting more swap. I wonder how many of the people having problems are running programs that need RAm way beyond what they have.

Yes, the 16gb model is running with better breathing space and not reliant on Swap so soon. With only Activity Monitor open 5.99gb already used which leaves very little remaining RAM for any applications I'd want to open on a 8gb model.

 

[Maximara]

Yes, the 16gb model is running with better breathing space and not reliant on Swap so soon. With only Activity Monitor open 5.99gb already used which leaves very little remaining RAM for any applications I'd want to open on a 8gb model.

Another factor is the size of the SSD. Quite frankly I don't know why in this day and age anyone would be happy with 256 GB.

 

[sneeks]

Another factor is the size of the SSD. Quite frankly I don't know why in this day and age anyone would be happy with 256 GB.

If funds had allowed I would have preferred a larger SSD but for my use I have plenty available and I use a external SSD to store the majority of my files. 256gb works for my needs. Finding the extra money for the 16gb was my main priority. The 8/512gb and the 16/256gb models were both priced at £759 at the Apple refurbished store. With all my apps installed I have 180gb available on the SSD.

 

[telo123]

Another factor is the size of the SSD. Quite frankly I don't know why in this day and age anyone would be happy with 256 GB.

For the general public (such as most students), 256 is more than enough. Quite frankly, I dealt with 128 GB 2015 rMBP for a couple years before I got my M1 Air, relying primarily on cloud storage.

 

[Snowii]

Yes these are all from my 16gb Mini, running these on the 8gb caused swap to be used but unfortunately that’s now gone so I cannot show screenshots of that. RAM usage was the same on the 8gb so did require Swap.

Sneeks - just going back to lightroom, how does the mini behave let's say after an hour of development work swap-wise?
From other sources it appears that lr caches quite a bit and uses swap fairly lot no matter the ram size, it's good to see no swap when you load it but real usage numbers may be way different. Thanks!

 

[JamSandwich]

Another factor is the size of the SSD. Quite frankly I don't know why in this day and age anyone would be happy with 256 GB.

I'm not sure anyone is happy with 256GB (who doesn't want some more space?), but I just picked up a refurb Mini and for the price Apple charges for storage, I'm happier to find a much more cost-efficient external solution.

(On the original topic, FWIW, on Day 2, total GB written on this machine were 289, which includes whatever happened before the machine reached me, setup and a small amount of web browsing and me fumbling to install smartmontools. I haven't done much with it yet, but it's increased only marginally from there)

 

[sneeks]

I'm not sure anyone is happy with 256GB (who doesn't want some more space?), but I just picked up a refurb Mini and for the price Apple charges for storage, I'm happier to find a much more cost-efficient external solution.

(On the original topic, FWIW, on Day 2, total GB written on this machine were 289, which includes whatever happened before the machine reached me, setup and a small amount of web browsing and me fumbling to install smartmontools. I haven't done much with it yet, but it's increased only marginally from there)

My refurbished Mini SSD data written increased from 345gb to 562gb but has now stabilised, i'm assuming because iCloud has now fully synced with the Mac.

 

[sneeks]

Sneeks - just going back to lightroom, how does the mini behave let's say after an hour of development work swap-wise?
From other sources it appears that lr caches quite a bit and uses swap fairly lot no matter the ram size, it's good to see no swap when you load it but real usage numbers may be way different. Thanks!

I’ll post a update here when I’ve had the chance to test this fully for myself.

 

[robfoll]

Ouch!

What is your typical usage? Which applications? which browser and how many tabs?

How much swap space do you typically use? I found that this was the biggest driver of high disk write volumes.

Moving to Edge browser with either the automated timed tab suspension (set to 30 minutes without use), or using the "Tab Discard" extension made a huge difference to my memory usage compared to Safari, and this was reflected in much smaller disk write volumes (from c. 500GB/day to 50-100GB/day).

Upgrading to MacOS 11.3 also may have helped somewhat.

Nothing much, all-day usage of Mail, Safari and Daylite CRM, a small amount of InDesign & Photos. Safari could be 5 to 20 tabs max.

Get an external thunderbolt enclosure and a m.2 ssd.
Mine is barely slower than the internal drive and it's easy to replace when you're 'out-of-writes'.
Might set you back $200-$300, but you'll never have to worry about this again.

Not very feasible on an out & about laptop and sure as hell should not be necessary. Cheers Robert

 

[Fomalhaut]

Nothing much, all-day usage of Mail, Safari and Daylite CRM, a small amount of InDesign & Photos. Safari could be 5 to 20 tabs max.

Not very feasible on an out & about laptop and sure as hell should not be necessary. Cheers Robert

Have you tracked your application memory usage, swap used, and disk writes throughout the day after a reboot?

Without knowing the numbers and the trends, it's very hard to determine the cause of the issue.

Do you have any other disks involved? TimeMachine external / network drives? NAS?

There are all sorts of things to try, mentioned in this thread. First stop is to try to identify memory usage and cause of disk writes. There are command line tools, but start off with the Activity Monitor data and post back here.