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
This is the question I would like to see an answer to!
One possibility is that SSD overprovisioning (which could be up to about 25% - probably less in a consumer machine) has already been used up, that the 2% represents the "tip of the iceberg", i.e. all the 25% overprovisioning has been used up, and I'm now "eating" the cells in the specified capacity (1TB). That might imply a TBW of 800TB and up.
If this were the case, I would expect to see that 2% figure increase much more rapidly. I'll keep tabs on the number.
Last edited:
How long have I spent poring through this thread? All fascinating it certainly is, and provoked me to find out how to install command line tools, smartmon and suchlike. For anyone else interested but slightly baffled the instructions I followed are here.
I received my M1 Air 512GB on January 5th and am a very light user because I use my iMac all day and maybe use the Air for an hour or so in the evening. My results:
Almost a terabyte written, apparently, but however I look at things the computer will outlive me, so I'm just not concerned. Also I have a feeling that Apple's engineers would (probably 😀) not have made a colossal error resulting in premature SSD failures, for it is SSD failures people are concerned about, isn't it? Otherwise, who really gives a hoot about the numbers? If the AS computers have higher disk write activity than Intel systems is that of itself important? It does remind of how I miss the disk activity LED of yesteryear, but why should it matter unless there are actual SSD failures occurring?
Just the thoughts of one user who's still very happy with his M1 Air.
I received my M1 Air 512GB on January 5th and am a very light user because I use my iMac all day and maybe use the Air for an hour or so in the evening. My results:
Almost a terabyte written, apparently, but however I look at things the computer will outlive me, so I'm just not concerned. Also I have a feeling that Apple's engineers would (probably 😀) not have made a colossal error resulting in premature SSD failures, for it is SSD failures people are concerned about, isn't it? Otherwise, who really gives a hoot about the numbers? If the AS computers have higher disk write activity than Intel systems is that of itself important? It does remind of how I miss the disk activity LED of yesteryear, but why should it matter unless there are actual SSD failures occurring?
Just the thoughts of one user who's still very happy with his M1 Air.
The bolded text is pretty close to my experience too. I can see a lot of write to the swap files once swap gets into the 4-5GB region. When I had 10-15GB swap in use (with c. 95 browser tabs opens), I saw 2.6TB of writes over 3 days, i.e. circa 850GB/day. I closed browser tabs pronto after seeing that.
Yesterday I had about 180GB of writes. Previous days have been 50-100GB. I'd like to keep writes within that lower limit on average.
Currently, my average (TBW/power-on-hours) is 23.9GB/hr. I hope changing my usage pattern (closing apps, and especially keep browser tabs to a minimum) will reduce that average to around 10GB/hr. I would consider that acceptable.
I still use a 2011 Mini (with retrofitted SSDs) and I would expect any machine to still be functional after 15 years, and to have at least 8-10 of usable service life (i.e. software support and "acceptable" performance with current software). Typically I don't keep my machines that long, but donate them, but would still like to think they are in service for many years after I dispose of them.
im at 20tbw from nov 18 to now , i find these numbers crazy.
while not a direct comparison, i have a 1tb ssd nvme WD black in my unraid server running 8 vm's , several windows 2016 servers and windows workstations that are used almost daily. it only has 120tbw. over almost 2 years.
while not a direct comparison, i have a 1tb ssd nvme WD black in my unraid server running 8 vm's , several windows 2016 servers and windows workstations that are used almost daily. it only has 120tbw. over almost 2 years.
You do realize that those numbers are nearly identical? 20 TBW since Nov 18 would be 111 days which works out to 180 GB/day. And 120 TBW over 730 days is 164 GB/day. So what's crazy?
Its sad that after 43 pages of valuable research, experiments and discussion, apple still has said nothing about this. I cannot wait that much longer to buy a MBA M1.
Many including people here say this only affects a small group of people, so.....should I just go ahead and buy it? 😆
Many including people here say this only affects a small group of people, so.....should I just go ahead and buy it? 😆
A car owner would be extremely happy with an Alfa Romeo or a Jaguar's reliability if it stays parked in a well ventilated pristine garage and goes out once a month for an occasional breakfast with the grandma in downtown or a beach side restaurant.
Your usage is not typical - based on your description 2 hours of use max per day - 2months = 120 hours or use with very light browsing. Sounds like an iPad usage to me - perhaps atypical for an average macbook user. You mentioned the workhorse is an iMac, so really your numbers can't be used to justify what we are seeing. For most people the compute power of the M1 and the low power/thermals makes it a better workhorse for real work.
Mine chewed up 2.12 TB in 13 days enough for me to be alarmed resulting in a return label getting printed.
Enjoy your M1, its a beautiful machine that may last you quite a while and hopefully outlives your iMac!
Last edited:
I'd say that would be a safe bet. If it is a serious issue (i.e. with threats of having serious percentage of premature failures and claims against them / class-action suits / product recalls) then they will pull out all the stops to fix this issue.
I suspect the issue is not as serious as it sounds for nearly everyone. I would expect some sort of technical statement form Apple about the cause of high write volumes with suggestions for reducing it.
Sounds like you are suggesting DiskGate? I doubt it - mostly not gonna happen - too many 3rd party apps that could be blamed by Apple and we would see some proverbial :shrugs: . And if it comes - its gonna take months!
At the end of the day its like Covid. Total Disk Failure percentage may be small - but when you get affected its 100% for you. To each their own, not all machines are behaving the same way either, the mystery remains after numerous discussions and data logs.
And this will go in history books for M1 if acknowledged. Mx is on the way, OS or Hardware architecture whatever it is.
Last edited:
Everyone has been working with the idea that the SSD degradation is linear so this really doesn't explain things.
It certainly doesn't explain why the TB*100/percentage = TBW crosscheck produces at worst 6827 (170*100/2.49) TBW for the drive.
SSDs and SMART Data goes into the details of how things are supposed to work. 11 Myths About SMART Monitoring and SSD Data Protection shows a serious issue with this whole thing:
2. Standard SMART utilities like Smartmontools and CrystalDiskInfo adequately manage SSDs.
While those utilities may be adequate for the casual user, most such utilities only extract attribute data and don’t sufficiently parse or interpret the data. In other words, they don’t necessarily provide immediately usable data.
Thes easiest way to test smartmontools is to run it on a Windows Machine with an SSD of a known TBW and see if the back calculation agrees with that or produces some off the wall result. If the later then the tool is giving us garbage data.
I don't know if this is useful information or not, but my 2015 12-inch MacBook, Apple's oldest soldered down NVMe device that I have used practically every day since November 2015 only reports 7.9TB of writes.
Granted nobody in their right mind is doing intensive work on such a device but I have at times managed to redline the memory pressure.
Granted nobody in their right mind is doing intensive work on such a device but I have at times managed to redline the memory pressure.
There is logic here. This anomaly has been seen on only a small number of units and interestingly in both M1 and Intel Macs. Maybe what is going on is since both use Apple custom designed controllers the SMART attribute numbering is not what smartmon is expecting? The Crucial article is clear that there is no uniformity among manufacturers about attribute codes.
Depends on what kind of user you are. If you do any kind of intensive work, get 16GB.
Also, don't be tempted to buy one just because of the hype. If you don't immediately need a new computer, it's more sensible to wait.
I'm the idea this behavior is due to the double SSD + unified memory of the ARM Mac. It would be curios to see the kernel_task usage of the guys who has multiple SSD inside the Mac Pro.
No, alredy discussed.
Update for my MBA 8/256 since 20th December 2020
Bash:
SMART/Health Information (NVMe Log 0x02)
Critical Warning: 0x00
Temperature: 20 Celsius
Available Spare: 100%
Available Spare Threshold: 99%
Percentage Used: 0%
Data Units Read: 6,526,154 [3.34 TB]
Data Units Written: 4,725,365 [2.41 TB]
Host Read Commands: 95,938,739
Host Write Commands: 40,185,195
Controller Busy Time: 0
Power Cycles: 224
Power On Hours: 40
Unsafe Shutdowns: 13
Media and Data Integrity Errors: 0
Error Information Log Entries: 0Yes. The fact that is effecting such a small number of people says something in of itself - whatever causes this is rare. The fact that for those over 0% se are seeing totally off the wall number when we plug the results in the the crosscheck formula TB*100/percentage = TBW.
I would “Unfollow” this thread before you do, because while there is some merit to some of the reports, many others seem to have descended into an obsessive mess.
Comparing the swap data of a 5+ years old Mac and MacOS (or OSX as it was called then) with a modern machine with the latest version of MacOS doesn’t make any sense to me.
Well, hopefully such a comparison would show that with gradual improvements of SSD storage media over time, brand new SSDs can be used more aggressively without wearing them out prematurely, no?
That would be true if the tools we were using weren't producing numbers that when you plug them into the simple formula: TB*100/percentage = TBW spit back up off the wall results
The fact that in one case we are getting an utterly insane 6,000+ TBW using the worse case number and a 8,000+ TBW using the numbers (170 TB for 2%) something is wonked. Yes, SSDs have improved but I seriously doubt they have improved that much.
Last edited:
the fact that 8 vm's are using the same amount as 1 instance of an os.
thats crazy.
did you just ignore completely that it was 8 different machines ? if it was one yes there would be little difference. but were talking 8 vm's.
That is exactly part of my point.
The other part is that different OSs behave in different way, especially after the file system has radically changed to APFS.
I see people saying "my old Mac xxx has done some many TBW in so many years, whereas my new M1 has done so many is so many weeks".
That type of comparison is pointless and doesn't prove anything. It doesn't compare like with like.
Different OS, different file system, different SSD generation.
A car owner would be extremely happy with an Alfa Romeo or a Jaguar's reliability if it stays parked in a well ventilated pristine garage and goes out once a month for an occasional breakfast with the grandma in downtown or a beach side restaurant.
Your usage is not typical - based on your description 2 hours of use max per day - 2months = 120 hours or use with very light browsing. Sounds like an iPad usage to me - perhaps atypical for an average macbook user. You mentioned the workhorse is an iMac, so really your numbers can't be used to justify what we are seeing. For most people the compute power of the M1 and the low power/thermals makes it a better workhorse for real work.
Mine chewed up 2.12 TB in 13 days enough for me to be alarmed resulting in a return label getting printed.
Enjoy your M1, its a beautiful machine that may last you quite a while and hopefully outlives your iMac!
I completely accept my use case is atypical, and I realise there is something a bit squiffy about M1 SSD usage.
For comparison, and just for jollies, this morning I ran the smart utility on my iMac. It tells me 14.5TB has been written since I received it on 1st June 2019, equivalent to ovewriting the 512GB SSD twenty-eight times. Over those twenty-two months the iMac has been on every single day from nine to six, writing an average of 0.65TB per month to the SSD. I do use Time Machine on the iMac but to an external SSD, Spotlight is enabled, I do not use any of the resource-heavy apps mentioned in this thread. The write figures seem high but I'm comfortable with them because I'm not expecting to keep the iMac for decades even if I am spared; I'm 73 and hoping to enjoy another couple of Mac generations. Similarly, my use case means my M1 should more than achieve the life I'd like it to reach, perhaps four or five years. Just as I refuse to be concerned about my iPhone's battery degradation I'm prepared to accept my SSDs are being devoured by time. That works for me, but if Apple can't reduce M1 SSD writes then I guess those who use them for real work have a bit of a problem.
I am now at quota 155 TB written in 73 days. Not good at all
Yesterday 1.538 TB were written in just 22 hours and 44 minutes. Unbelievable.
Edit: Swap is now at 8.3 GB.
Well, it's pretty easy to write that much with swap. If your apps are busy, the swap has to be paged-in and out constantly.
To whoever said that M1 does some "magic" and much RAM isn't needed: You have no clue what you're talking about.
Yes: The unified memory is fast and has performance advantages especially for GPU stuff. But it has no effect on quantity. Fast RAM is good. But the most important thing about RAM is that you have enough to fit all your applications and their data into it.
The reason why a 8GB M1 performs similar to a 16GB Intel is, that the M1 machines use much faster SSDs so swap has less performance impact. But it still shreds the SSD if you constantly use too much swap.
If all your apps you use simultaneously need more than, say, 8GB RAM, then buying a Mac with 8GB was a bad choice.
I'm pretty sure, that most of the disk writes come from swap on machines with not enough RAM for all the Apps that are used. Especially "modern" Apps are often Electron based. Those app easily eat 1-2GB of RAM (apps like Discord, VS Code, Slack, WhatsApp for Mac, ...).
The OS also needs RAM for its services and also for caches and buffers. macOS wants at least 4GB RAM for itself.
Yeah, your SSD will be shredded by the swap.
In your special case: 155TB / 73 days = 2.12TB/d
Which is pretty crazy though. It will take you 470 days to write 1PB of data. So at that rate your SSD might last 2, maybe 3 years depending on capactiy.
You also have to take wear leveling and write amplification into account. The fuller your SSD is the more writes it will actually do to the NAND. Even if you write just 1GB of data, your SSD might actually writes 4GB of data, because it has to move data before.
Now you can roast me for saying that the M1's RAM is not so magic as some people tell you...