ssd swap - high usage of Terabytes Written

[Robospungo]

That is only part of the calculations. That 19.7 is 1%. While that could be anywhere between 0.5% to 1.499..% we'll go with the 1%. 19.7 x 100 is 1900 TBW for 100%. Even if throw that out as ridiculously high we still have 1% in 49 days which results in 100% for 4800 days or 13.42 years

By your own numbers the drive has a lifespan of 13.42 years. Even if we go by the 800 TBW that resulted from a bank doing something very very silly with an entry level machine you are still talking about ~1989 days (800/(19.7/49)/365) or 5.45 Years.

Please explain to us just how long do want the SSD to last if 5.45 or perhaps 13.42 years are not enough. 😱

If I'm paying over $1200 (Canadian) for a laptop, I damn sure expect it to last longer than 5 and a half years. If the SSD were replaceable, this would be a different conversation.

 

[leons]

delete

 

[Fomalhaut]

The TBW value is measured but the Percentage Used is estimated by the vendor, in this case Apple. Measurements can be tested by comparing known write amounts with the value returned. I've done this and so have others. The measured amount of data written appears to be correct. On the other hand, if Apple made a mistake in any of these values, the most likely place would be in an estimate. Don't you think?

Exactly. I'm really not sure how much we can trust the "Percentage Used" figure. From the smartctl "man" page:

ssd - [SCSI] prints the Solid State Media percentage used endurance indicator. A value of 0 indicates as new condition while 100 indicates the device is at the end of its lifetime as projected by the manufacturer. The value may reach 255.

It's an estimate generated by the manufacturer. We don't know how Apple (or their SSD sub-contractor) has configured this. It may be optimistic.

 

[Fomalhaut]

Just to add some more data
3 months , 6.59TBW
0% used

Usage is limited to mostly native programs with the notable exception of MS teams .
Majority of usage is productivity apps/ Zoom/ Teams and of course, browsing (99% Safari)

On an average, I see around 70-90GB written every day I use the system for a full day .. and 20-40GB on days of intermittent usage

Now based on the above it would appear at first glance that the endurance of the drive is at least 660TB or more
And thus I should have no reason to worry about it as it would be at least 50*3 = 150 months before the drive crosses 50% use (50% being an arbitrary threshold beyond which I may need to look at disposal/ replacement(

Yet I keep looking at these stats because I am not sure if the usage stats are linear - or if they start rising only after the spare/ redundant blocks on the SSD are used up .
In other words, if the latter is true then the change from 0 to 1% will take a long time but will rise a lot more rapidly after

I have exactly the same question. Some SSDs have over 25% spare capacity to allow for faulty storage cells to be replaced without reducing specified capacity.

I could see a case where this "invisible" capacity is being used up, at say 2% (of the total disk) per month, with the "percentage used" at 0% for the whole year, and then suddenly after a year or so you start seeing the "visible" percentage use increase at 2% per month.

 

[leons]

Re-read the quote and my answer and thought it was a bit snarky! I "interpreted" the "between 60 and 150TBW", but it doesn't say that. It says a rating of 60 (what?) and 150TB - which could imply the "60" referred to some other metric. My bad!

It's obviously 60 TBW.... the same way "between 50 and 70 years old" is unambiguously referring to the range of the same unit (years).

No offense taken. It's "probably" as you say, and likely just poor writing by the author. We're all just trying to solve the problem here.

 

[leons]

And this uncertainty about the validity of the smartctl numbers is part of the problem.

My MBP16 with its 172TB written @2% used according to smartctl gives an unlikely results for the maximum TBW. We can't discount that it may be wrong and that the percentage is being incorrectly calculated. Or if it means something other than described in the documentation.

We can't necessarily trust any of the estimates for the estimated duration of the SSD if there are experimental observations that cast doubt on the numbers.

But you can see the numbers right in Activity Monitor. Those are correct unless you have more than one disk per "the professor" in the aforementioned video. In the comments, he even advises one commenter that 100DWPD is concerning. Mine (and many others) are way above that.

 

[leons]

We have one significant report from the video comments correlating use with wear:
"My 78 day old 256 GB M1 Air is writing 1.3 TB/day with non-professional, personal usage. The percentage used is 6%. I guess I do have the problem."

 

[leons]

I am now suspecting that MY problem is related to my excessive use of Safari Tabs (20-45 sometimes). Is there a way to tell Safari not to cache?

 

[Argon_]

I have exactly the same question. Some SSDs have over 25% spare capacity to allow for faulty storage cells to be replaced without reducing specified capacity.

I could see a case where this "invisible" capacity is being used up, at say 2% (of the total disk) per month, with the "percentage used" at 0% for the whole year, and then suddenly after a year or so you start seeing the "visible" percentage use increase at 2% per month.

Built in overhead to prevent capacity loss. I figured there'd be something like this, otherwise you'd see cap decreases relatively quickly. Thank you for the info.

 

[Fomalhaut]

I am now suspecting that MY problem is related to my excessive use of Safari Tabs (20-45 sometimes). Is there a way to tell Safari not to cache?

I typically used to have 80-100 Safari tabs open on my work computer, and at the start of this thread I saw that my SSD usage for several days was c. 850GB written per day.

I am now trying to close Safari tabs (or add to the reading list) more proactively to keep the open tabs down to <40-50. I have noticed an improvement in SSD writes; I'm now at 50-250GB / day.

Whether this is specific to Safari or just general usage of Swap I don't know. Safari tabs were the largest memory hog, and reducing the number of open tabs reduced my swap to healthier numbers.

I have found that when swap is >5GB (I have a 16GB Mini), SSD writes increase significantly.

 

[Fomalhaut]

But you can see the numbers right in Activity Monitor. Those are correct unless you have more than one disk per "the professor" in the aforementioned video. In the comments, he even advises one commenter that 100DWPD is concerning. Mine (and many others) are way above that.

I think "DWPD" means "Drive writes per day", which if I have understood correctly means the total capacity of the disk, written per day. So if you have a 500GB disk, 10 DWPD would be 5000GB (5TB).

I have seen specs for enterprise SSDs that have 3.0 DWPD rated for 5 years, so your 500GB disk would be writing 1.5TB per day, for up to 5 years, beyond which the drive is out of spec (& guarantee) and could fail.

100 DWPD would be a huge number that is well beyond the ratings of normal SSDs

Here's an ad for a Samsung enterprise SSD. It's rated for 5 years or 1366TBW. Considering the capacity is 960GB and its TBW rating, we could calculate its DPWD (for 5 years) as: (1366000/960)/(5*365) = 0.779.

So you can't even write the whole disk capacity once per day and still have an in-spec disk after 5 years.

Buy Samsung 983 DCT 960GB 2.5in U.2 Enterprise SSD for Business | Hard Drives & SSDs | Scorptec Computers

Samsung 960GB SSD, 983 DCT Enterprise, 2.5" 7mm U.2 NVMe PCIe, Max R/W 3300/1300 MB/s, V-NAND 3bit MLC, 440K/46K IOPS, 1366TBW, MTBF 2M Hours

www.scorptec.com.au

More information on DPWD here:

Home Page - Western Digital (WD) Corporate Blog

Home Page - Western Digital (WD) Corporate Blog

blog.westerndigital.com

 

[leons]

I think "DWPD" means "Drive writes per day", which if I have understood correctly means the total capacity of the disk, written per day. So if you have a 500GB disk, 10 DWPD would be 5000GB (5TB).

I have seen specs for enterprise SSDs that have 3.0 DWPD rated for 5 years, so your 500GB disk would be writing 1.5TB per day, for up to 5 years, beyond which the drive is out of spec (& guarantee) and could fail.

100 DWPD would be a huge number that is well beyond the ratings of normal SSDs

Here's an ad for a Samsung enterprise SSD. It's rated for 5 years or 1366TBW. Considering the capacity is 960GB and its TBW rating, we could calculate its DPWD (for 5 years) as: (1366000/960)/(5*365) = 0.779.

So you can't even write the whole disk capacity once per day and still have an in-spec disk after 5 years.

In this context, DWPD is the maximum writes per day that will keep you within the warranty/normal/expected specs of the drive. It is defined as .3 x Drive Capacity. So, for a 1TB drive, you wouldn't want to go above 300MB/day

 

[Fomalhaut]

In this context, DWPD is the maximum writes per day that will keep you within the warranty/normal/expected specs of the drive. It is defined as .3 x Drive Capacity. So, for a 1TB drive, you wouldn't want to go above 300MB/day

I imagine that the proportion of drive capacity (e.g."0.3 x drive capacity") would depend on the guaranteed specs of the drive.

There are probably differences between the expected life and the actual warranty period. The latter will be as low as local laws (and market positioning) as the manufacturer can get away with. The former is hopefully a lot longer!

I would be pretty disappointed with any computer that wasn't designed to last for 10 years of "normal" use, even if the component spec limits were necessarily lower to allow for the usual bell-curve of failures.

0.3 x drive capacity is quite low compared to the numbers people are quoting, but I can believe it on a consumer device if we are seeing "enterprise" drives with only 0.8 x drive capacity (for 5 years).

This would mean only 150GB written on my 500GB SSD per day, which is close to my average usage. That's less than 200TB after 5 years. It all depends on the TBW rating of the drives. If it is 600TBW, that would be 15 years....acceptable in my opinion.

 

[leons]

Yes. The problem is, that for a 256GB drive (like mine), the .3 "normal" comes out to be about 77GB/day. I (and others) are writing 10 times that, which by any standard is well beyond expectations for any consumer SSD.

 

[Fomalhaut]

Yes. The problem is, that for a 256GB drive (like mine), the .3 "normal" comes out to be about 77GB/day. I (and others) are writing 10 times that, which by any standard is well beyond expectations for any consumer SSD.

Yes, indeed. I doubt that Apple SSDs are radically different from other high-end models made by Samsung, WD, Sabrent etc. We might see 500-800TBW for a 250GB drive if we are lucky, but it could be less. At the low end, even writing 250GB/day could mean the SSD is only in-spec for 5 years. If the TBW is less, or you really do write >500GB/day consistently, then there could be a real problem with the longevity of Macs with this issue.

 

[leons]

Yes, indeed. I doubt that Apple SSDs are radically different from other high-end models made by Samsung, WD, Sabrent etc. We might see 500-800TBW for a 250GB drive if we are lucky, but it could be less. At the low end, even writing 250GB/day could mean the SSD is only in-spec for 5 years. If the TBW is less, or you really do write >500GB/day consistently, then there could be a real problem with the longevity of Macs with this issue.

Unfortunately, it's real.These numbers are not from some "tool", but directly from Activity Monitor. And it is a MacBook Air 8/256 purchased 10 days ago, which has never had another disk connected in any way. :-(

 

[Fomalhaut]

Unfortunately, it's real.These numbers are not from some "tool", but directly from Activity Monitor. And it is a MacBook Air 8/256 purchased 10 days ago, which has never had another disk connected in any way. :-(

View attachment 1748315

That's scary! How much swap space are you using (from Activity Monitor)?

You can see what a process is writing data to with this command (e.g. for the kernel_task):

sudo fs_usage -w -f diskio kernel_task

I haven't seen any specific app or usage pattern that triggers large disk writes other than it appears to happen a lot more when I'm using significant (> 4-5GB ) of swap. I see lots of lines like this:

13:51:47.473336 PgOut[AP] D=0x035cdf5e B=0x40000 /dev/disk3s6 /System/Volumes/VM/swapfile6 0.000003 W kernel_task.1523

I'm glad I got the 16/512GB model though....the combination of having a lot more disk swapping (& hence writing) on the 8GB model (or having a smaller SSD with a lower TBW) would have been a mistake for my usage.

 

[antwormcity]

I am now suspecting that MY problem is related to my excessive use of Safari Tabs (20-45 sometimes). Is there a way to tell Safari not to cache?

Try with Chrome, Safari chewed up my disk on an M1 Air, Chrome will hog memory as we all know, so I thought and was using Safari all along on a new Air. However, do that experiment and perhaps see. My TBW was around additional 1TB while using Safari and during similar time duration it went to half of that with switch to Firefox/Chrome.

 

[Ningj]

I typically used to have 80-100 Safari tabs open on my work computer, and at the start of this thread I saw that my SSD usage for several days was c. 850GB written per day.

I am now trying to close Safari tabs (or add to the reading list) more proactively to keep the open tabs down to <40-50. I have noticed an improvement in SSD writes; I'm now at 50-250GB / day.

Whether this is specific to Safari or just general usage of Swap I don't know. Safari tabs were the largest memory hog, and reducing the number of open tabs reduced my swap to healthier numbers.

I have found that when swap is >5GB (I have a 16GB Mini), SSD writes increase significantly.

Website memory usage, no matter the browser is getting extreme. Just have a look in activity monitor and filter processes for your respective browser. e.g. I've got 5 tabs open, this forum, FB, YT, gmail, and a local news site. Thats 1.2G just for 5 tabs! Imagine what 40-50 tabs might look like

 

[leons]

Try with Chrome, Safari chewed up my disk on an M1 Air, Chrome will hog memory as we all know, so I thought and was using Safari all along on a new Air. However, do that experiment and perhaps see. My TBW was around additional 1TB while using Safari and during similar time duration it went to half of that with switch to Firefox/Chrome.

Mark me as really PISSED! Before I bought my Air, I was using a Chromebook. I read EVERYWHERE that Safari was optimized for Mac, and that I needed to give up my memory-hungry, non-performant Chrome. So, for 10 days I try to reproduce the functionality I had with Chrome, plugins, etc. in Safari. And guess what? Over the 10 days, the uber-efficient and integrated Safari eats up 8TB of my SSD. So, I do a test (just now). I take a look at "Memory Pressure" with my 39 Safari Tabs open. It is SO RED! So, I close Safari and open up Chrome with 39 tabs (one of which I am typing in now). Guess what? Memory Pressure couldn't be greener. WTF, is Safari THAT f***** up?

 

[Tev11]

Mark me as really PISSED! Before I bought my Air, I was using a Chromebook. I read EVERYWHERE that Safari was optimized for Mac, and that I needed to give up my memory-hungry, non-performant Chrome. So, for 10 days I try to reproduce the functionality I had with Chrome, plugins, etc. in Safari. And guess what? Over the 10 days, the uber-efficient and integrated Safari eats up 8TB of my SSD. So, I do a test (just now). I take a look at "Memory Pressure" with my 39 Safari Tabs open. It is SO RED! So, I close Safari and open up Chrome with 39 tabs (one of which I am typing in now). Guess what? Memory Pressure couldn't be greener. WTF, is Safari THAT f***** up?

There’s definitely a bug in Safari. It typically caches websites and preserves it in RAM and swap memory. You can’t really disable caching in Safari by setting it’s cache folders to read-only — unless I did it wrong (by maybe setting a necessary folder to read-only), Safari kept using the physical memory instead and not used SWAP. Obviously, that had caused the system to freeze as it ONLY used the physical RAM.

Chrome on M1 Mac’s don’t seem to use up that much RAM (at least per my own usage). By disabling Cache in Chrome (by setting its cache folders to read only), you avoid high disk writes when streaming YouTube videos. Disabling Cache on Chrome still allows it to use virtual memory.

However, I still use Safari because of iCloud Keychain and the irreplaceable Reader Mode it has. Safari is the only browser that Netflix permits 4K streaming on macOS — all others are limited to 720p.

Note: some websites you might be using might be poorly coded as well. Some website developers refuse to code their website to work well on WebKit (engine that Safari uses). Poorly coded websites might lead to higher than normal usage of system resources.

 

[Maximara]

Try with Chrome, Safari chewed up my disk on an M1 Air, Chrome will hog memory as we all know, so I thought and was using Safari all along on a new Air. However, do that experiment and perhaps see. My TBW was around additional 1TB while using Safari and during similar time duration it went to half of that with switch to Firefox/Chrome.

Careful there. Chrome will generate little helpers which can be easily missed in Activity Monitor giving the impression Chrome is writing far less data to the drive than it actually is.

 

[Maximara]

Website memory usage, no matter the browser is getting extreme. Just have a look in activity monitor and filter processes for your respective browser. e.g. I've got 5 tabs open, this forum, FB, YT, gmail, and a local news site. Thats 1.2G just for 5 tabs! Imagine what 40-50 tabs might look like

View attachment 1748323

Ouch. I knew the web had become a memory intensive mess but that is insane.

 

[Spindel]

Unfortunately, it's real.These numbers are not from some "tool", but directly from Activity Monitor. And it is a MacBook Air 8/256 purchased 10 days ago, which has never had another disk connected in any way. :-(

View attachment 1748315

Any external drives or network volumes connected?

 

[Maximara]

There’s definitely a bug in Safari.

It appears to not be releasing its RAM (memory leak I think). I was at 80% of my 8GB (2013 Intel Mac) which five tabs open, one on Youtube. I quit Safari and reopened it with all the tabs still open and I am now using 48%. Safari has other issues. It will generate weird graphic errors on Nvidia Macs (though this appears in other programs as well)