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
So, nothing to do with the M1 chip then. But thanks for sharing.
BTW, if you are constantly downloading stuff well you can have a lot of writes.
Oh, but it is very quiet lately.
The "issue" (there is no consensus about whether there is actually an M1-related issue yet) has gone from rare to extremely rare.
Thanks for showing up, though. It feels lonely here, sometimes.
OK I'll bite. As you can see from my earlier posts here I am one of those affected by the excessive writes to the SSD.
A bit of background:
8GB/256GB M1 MBA
No photo editing, no Video editing, just 2 hrs Zoom/month, no builds no tools installed, I just use FireFox with the Auto Tab Discard, on 11.5.1.
Someone suggested (and some others rubbished) the idea of contacting Apple about this. Well, I thought why not? If there eventually proves to be a problem then it is on record that I enquired. So on chat with 'Shell', she (?) suggested that resetting the following:
I'm told the way to do this is to power down, close the lid, and wait a minute or so. So I did that but no clue as to how often.
I also followed up on this:
Looking for some idea about the resilience of the SSD a figure of 200TBW came up. So my disk(s) - there are 2 physical SSDs each 128GB - are through possibly a 1/4 of their lifetime warranty, say another 7*3 - 21 months. Let us say it is in practice double, so around 4 more years at this rate. Not good news.
The table below is a log of results of usage under different trials:
| 14/07/2021 | 95,161,423 | 48.70 | Logout/in | |
| 15/07/2021 | 95,773,524 | 49.00 | Logout/in | |
| 16/07/2021 | 96,931,715 | 49.60 | Logout/in | |
| 17/07/2021 | 97,107,583 | 49.70 | Logout/in | |
| 18/07/2021 | 97,507,781 | 49.90 | ||
| 19/07/2021 | 98,137,278 | 50.20 | ||
| 20/07/2021 | 99,130,295 | 50.70 | sleep only | |
| 21/07/2021 | 101,984,269 | 52.20 | sleep only | |
| 22/07/2021 | 105,292,928 | 53.90 | sleep only + charging | |
| 23/07/2021 | 105,427,457 | 53.90 | logout/in | |
| 24/07/2021 | 105,437,355 | 53.90 | logout/in | |
| 25/07/2021 | 105,491,531 | 54.00 | sleep only | |
| 26/07/2021 | 106,041,737 | 54.20 | sleep+power | |
| 27/07/2021 | 107,160,478 | 54.80 | sleep+power | |
| 28/07/2021 | 107,325,807 | 54.90 | sleep+power | |
| 29/07/2021 | 107,923,627 | 55.20 | 0.4846174129 | sleep+power |
| 30/07/2021 | 112,242,402 | 57.40 | sleep only | |
| 30/07/2021 | 112,269,066 | 57.40 | PoPo | |
| 31/07/2021 | 112,342,972 | 57.50 | sleep+power | |
| 01/08/2021 | 112,377,451 | 57.50 | sleep+power | |
| 02/08/2021 | 112,441,676 | 57.50 | sleep+power | |
| 03/08/2021 |
So you can see for some silly reason having the power connected makes the problem go away, to a lesser degree logout/login also helps a lot. Overnight 29 July it jumped 2.2TBW on sleep and no power cord connected!
So just as a matter of interest, do those on here generally keep the power connected at least overnight and are not getting the excessive write activity, please?
I would love to know what the difference is with those who do not experience this effect. I would also add that the people on this thread (and in other forums) discussing SSD writes must be a tiny tiny percentage of M1 Macbook users, but that doesn't mean that the other 99.99% of users are not experiencing this problem.
It's just that they haven't looked.
Last edited:
If my M1 Mac was was having such issue and unable isolate it to aberrant third party applications or correct with the likes of Onyx I'd back up my data and do a full restore;
Revive or restore a Mac with Apple silicon using Apple Configurator 2
In rare circumstances, Apple computers become unresponsive and the chip’s firmware must be revived or restored.
support.apple.com
Documenting the issue/concern is the smart thing to do. If the drive does fail prematurely the user will have something far more solid to fall back on, especially if the problem is tracked, documented and Apple updated periodically.
I have looked and there's no indication that my M1 MBP is writing excessively to the drive. I don't doubt some are incurring this issue as the data speaks for itself. Personally I don't see it as being widespread similar to the Butterfly Keyboard, Radeon dGPU etc. where a very significant number of Mac's were affected.
Q-6
The percentage used will give you an idea of the total lifespan of your 256 GB SSD. The numbers posted here suggest it is much, much greater than 200 TBW.
I almost never leave the power cable connected overnight, and I don't log out or power off either. I never had the issue, though, so I don't know if this helps you in any way.
I do this on all of my MacBook Pros. My daughter does it with her M1 MacBook Air.
Having your download folder on an old platter drive will save your SSD from this. There are dozons of articles on how to have a secondary user folder on an external drive. Just make sure you have one on your main system drive.
Where? When we have been given percentages and extrapolated them we either got TBWs in the low thousands and/or decades of future use. What I have said in much earlier posts:
For example, one person reported 1.03 TB for 8 hr in a month and another reported 4.3 TBW over 5 months (0.86 TB per month). When put in the context of The SSD Endurance Experiment: They’re all dead a linear progression is perfectly reasonable and taking the 600 TBW a bank using an M1 as a postgres server got (and there is something wonky about the numbers) that works about to 50 years for the first and 68 years for the second.
--
The quick and dirty TBx100/percentage (based on the 2014 test) results in 1745 TBW which given the range makes a 1600 TBW totally reasonable. More over if 4 months = 2% then based on the test 100% = 200 (4 x 50) months or 16.7 years. Even if we cut that in half it is still 8.3 years.
--
Just to clarify. The way the percentage works is it rounds to the closest integer. So 0.4999... rounds to 0% but 0.500...1 rounds to 1% So the quick and dirty formulas are:
14.85 x 100 / 0.4999 or ~2970 TBW for 100%
14.88 x 100 / 0.5 or ~2976 TBW for 100%
So 2900 TBW for 100% seems within reason.
As I said before people look at only part of the data and freak out when the other parts indicate something totally different.
--
That works out to a TBW drive total of 2500 for 100% or a life span of ~6.9 to 8.3 years assuming linear consumption.
--
That works out to be a range of 11,993 (assumes actual 1.5...1%) to 7160 (assuming 2.4999999...%) TBW before 100% is reached. I would go with the lower of those or ~7160 TBW
--
The worst case we have is a 518 GB drive going bye bye at the 600 TB mark and there are serious questions about those numbers (they don't jive with others reporting their usage) and the M1 Mac was used in something that is write happy.
The key number is not how much is written on its own but also Percentage used. A 0% means at worst only 0.49999.% of the SSD's rated lifetime capacity has been used. This gives us at worst 800 TBW before it hits 100% (and most SSDs even in 2014 went way past that number)
If 10 days gets 4 TB than it should take 2,000 days (800/4*10) to hit the 100% mark. That is ~5.48 years - more than enough time to get an external or replace the machine. And this is the worst case. May people are coming up with 1000+ TB for their drives.
--
And those were the one I could find after all this time...there are others where I took the numbers apart.
Last edited:
Here's my thoughts about it after acquiring this Air in May.
I'm a Cloud Engineer and I do a lot of work with HDB Studio, Terminal, Visual Code and a lot of web apps. My normal day I have 2/3GB of SWAP however on some heavy days I'm doing a lot of SQL scripting and running a lot of stuff I can reach 5-6GB. To mitigate this issue with this Rosetta translation app I moved my workflow to Citrix.
This is how it is since the end of May.
I'm a Cloud Engineer and I do a lot of work with HDB Studio, Terminal, Visual Code and a lot of web apps. My normal day I have 2/3GB of SWAP however on some heavy days I'm doing a lot of SQL scripting and running a lot of stuff I can reach 5-6GB. To mitigate this issue with this Rosetta translation app I moved my workflow to Citrix.
This is how it is since the end of May.
Sadly 4.79 TB at 0% doesn't tell us much. As I said before it could be 0.0000(...)01% to 0.499999...%. Until one hits at least 1% the DUW value doens't tell you anything really useful.
This was the list of reported results on June 12th
256GB Macs
Souko 1% 15.3TBW
Leons 1% 14.9TBW
Thistle41 1% 29.7TBW
The Synchroniser 2% 25.2TBW
VitoBotta 18% 268TBW
512GB Macs
Mike Boreham 0% 21.7TBW
Formalhaut 1% 25.9TBW
1TB Macs
jdb8167 0% 10TBW
Dieselm 6% 150TBW
Looks like crossover from 0-1% may occur about 12TBW on 256GB Macs and around 24TBW on 512GB Macs, but it is not an exact science.
256GB Macs
Souko 1% 15.3TBW
Leons 1% 14.9TBW
Thistle41 1% 29.7TBW
The Synchroniser 2% 25.2TBW
VitoBotta 18% 268TBW
512GB Macs
Mike Boreham 0% 21.7TBW
Formalhaut 1% 25.9TBW
1TB Macs
jdb8167 0% 10TBW
Dieselm 6% 150TBW
Looks like crossover from 0-1% may occur about 12TBW on 256GB Macs and around 24TBW on 512GB Macs, but it is not an exact science.
Very true as that "1%" could be anything between 0.5% to 1.49999...% If we take the worst of that range we get 100/1.49999 * 12 TBW or on insured 800 TBW for the 256GB Drives.
However there is a major glitch in the data:
Thistle41 1% 29.7TBW
The Synchroniser 2% 25.2TBW
Something is seriously wrong there as no matter how the percentage measures things there is no way 1% 29.7TBW and 2% 25.2TBW makes any sense, The 29.7 TBW value for the 256GB Drive should have a higher percentage not a lower one.
As I said quoting Conan-Doyle's creation "When you have eliminated the impossible, whatever remains, however improbable, must be the truth".
If there was a smoking gun that the data we are getting is questionable this is it.
Thanks, yes it helps in the sense that the power attachment status is not the defining root cause of the problem. As to why my particular setup is sensitive to that is a complete mystery. If you had said 'Yes, I always leave it connected overnight' I would be on to something.
Yes, I agree, and thanks for digging down into these data results. It gives some hope that the reporting is somehow not accurate.
Mine stayed at 1% up to 31.1 then next time I looked 2% at 37.2 then flipped over to 3% at 48.7.
Last edited:
You say this repeatedly and are corrected repeatedly. There is write amplification to consider which can totally explain the discrepancy. A nearly full drive will have more write amplification than a nearly empty one.
It’s as accurate as Apple wants it to be. I wrote a command line tool using only Apple supplied APIs that only tests against Apple designed SSD controllers and the results were exactly the same as the other open source tools and commercial applications.
smartTBW
No, not seriously wrong at all.
If I had said "the crossover from 1% to 2% happens in the range 25TBW to 30TBW" then Thistle and Synchronisers results do not look anomalous. 25-30 is not a very big error band at all for a subject like this.
Please note I did say that this is not an exact science. As @jdb8167 points out factors like write amplification will cause variability.
Last edited:
Thanks for extra data...I hadn't seen this when I replied to Maximara just now.
I don't think it changes anything except that the scatter for 1 to 2% crossover is larger than the 25TBW-30TBW I postulated which would not be at all surprising.
Well, this data shows that after 3 months of usage you have 0% percentage used. It means that in the worst case (the zero jumps to 1% in the next minute) your SSD will last for about 3 months times 100. Which is 300 months or 25 years.
Not bad, I think. The battery will probably die long before.
Just to be clear here....I am not trying to say there is an amazing correlation between TBW and % life used here. There is far too little data for that. I am simply rejecting that there is something seriously wrong with the data because two datapoints apparently contradict. The data available is not inconsistent for the issue.
In any case as others are saying, I think that the data is pointing to much longer lives than feared at one point, such that my concern level has largely gone, leaving curiosity, especially why a few machines seem to be an order of magnitude worse.
I've found DriveDx handy for keeping an eye on my disks (it gives you a figure for total data written), and you get a couple of weeks free trail when you download.
Incidentally, after a month my new M1 iMac is writing as near as makes no difference the same about of data to its SSD as my intel iMac.
Save your $ just use @jdb8167 Terminal tool as it will present exactly the same results as the likes of paid apps such as DriveDX, as they rely on the same API's.
Q-6