ssd swap - high usage of Terabytes Written

[Spindel]

I am 99% close to buying the MBA M1. Should I postpone my purchase then?

Its confusing because many are panicking but many are also saying not to worry...

No need to worry, if you need a new computer get it.

 

[Answer_i3]

There is some info: have same story about ssd samsung eve m2. After installing TBW was 64 Tb(about 2 month), then after research it turned out - iCloud wrote something with 200mb/sec...

 

[KUKitch]

Purchased mine in mid-December, and it is just a personal laptop so I imagine I don't see the levels of usage that many on here have. This is mine:

=== START OF SMART DATA SECTION ===
SMART overall-health self-assessment test result: PASSED
SMART/Health Information (NVMe Log 0x02)
Critical Warning: 0x00
Temperature: 27 Celsius
Available Spare: 100%
Available Spare Threshold: 99%
Percentage Used: 0%
Data Units Read: 6,633,999 [3.39 TB]
Data Units Written: 3,944,970 [2.01 TB]
Host Read Commands: 126,928,190
Host Write Commands: 45,108,084
Controller Busy Time: 0
Power Cycles: 169
Power On Hours: 54
Unsafe Shutdowns: 4
Media and Data Integrity Errors: 0
Error Information Log Entries: 0

 

[theluggage]

I am 99% close to buying the MBA M1. Should I postpone my purchase then?

Its confusing because many are panicking but many are also saying not to worry...

There's a risk with any new-ish machine. Don't be paralysed by that. However, it depends on what you personally deem acceptable.

In terms of this particular problem, it may be that the worrying readings are just wrong. At the very least, there's a huge amount of uncertainty in the figures and how they translate into life-time If not, the handful of worst cases (100-200 TBW in a couple of months) could, if real, kill the system in a couple of years - but they do seem to be extreme cases. People who are seeing ~10-30 TBW/month seem more common, and that shouldn't be a problem for 8 years or so... Others are seeing far less than that.

(Observation - people don't install software like Homebrew and smartmontools if they're just doing spreadsheets and updating TwitBook. The first reporters of this are likely not "typical users" - doesn't mean they're dishonest, just that their workloads aren't necessarily representative).

However, this still highlights a couple of home truths:

1. Soldered-in SSDs - one of the components most likely to fail - are a bad idea. Full stop. An SSD is a consumable item with a limited number of writes, so it should be replaceable. I don't think there's anything good about a system which will predictably be "beyond economical repair" within 10 years... and even if the current scare is false, it is still perfectly feasible that an OS bug, a rogue app or just very heavy swap use could greatly reduce that lifespan. The current lifespan of SSDs is "acceptable" but there's not sufficient headroom to justify making them non-replaceable.

Now, every other computer you might consider has other pros, cons and design flaws so that doesn't mean "never buy a Mac" but the soldered-in SSD is a "minus" point that you need to weigh in to your deliberation. Some people will be happy if it lasts 3-4 years (...although it might not keep its resale value after that the way past Macs have).

2. Ignore the nonsense that some people have posted about M1 RAM going twice as far as on Intel "because Unified Memory". Sorry - 1GB of data on Intel is 1GB of data. There are theoretical ways in which UMA might reduce RAM requirements for frame buffers etc. in specific, M1-optimised apps/workloads but nothing that's gonna make your RAM go twice as far - if your apps need to hold x GB of data in memory on Intel they'll probably need to hold x GB of data in memory on M1, and if your Mac doesn't have that much RAM it will start swapping to disc. Quite likely, the test being done on the Intel Mac wasn't limited by RAM in the first place (lots of videos failed to show the critical 'Memory pressure' reading) - or the M1 was swapping like crazy but it didn't show because of the fast disc I/O on the M1, plus the M1 being generally faster at some jobs (like video codecs).

Bottom line there - don't buy a MB Air to do the job of an iMac or 16" MBP with 32GB+ of RAM just because some benchmarks show it having comparable performance. "proper" 16"MBP/5k iMac replacements (hopefully with larger RAM options) should be along soon and if your workflow really is RAM intensive you ought to see a significant benefit.

Remember: the Air, the Mini and the 2-port 13" MBP are Apple's entry-level computers for basic office/communication/entertainment - it's great that they'll also cope with casual video editing/photo work/audio production/number crunching, but they're not for people who are doing those things 5 days a week. It's not rocket surgery that RAM-intensive work with insufficient RAM will hammer the SSD with swap usage. Again - up to you, depending in what you want the machine for.

 

[CMMChris]

Please, stop blaming swap usage for excessive SSD wear finally. Yes, it does put additional wear on the SSD but it's a non issue. I do a lot of editing (4k30, 4k60, sometimes 8k) and multitasking on my 16GB MBP and I am using the swap file a lot. But it's not like swap causes a TB of written data per day. The most I've seen was 12GB on a single day. The impact swap has on your SSD wear isn't that huge.

Here are some fresh stats from my M1 MBP. Not much has changed from the last time. Again: I am using that thing as my main productive machine since late November.

Blaming the swap for the TBW issue some users have is nonsense.

 

[daverdfw]

so just to report back after disabling spotlight things seem better, data written by kernel task is 205GB over almost 20 hours of uptime. Which is much better than what I was seeing before. the system also seems less laggy than it was before.

 

[flopticalcube]

so just to report back after disabling spotlight things seem better, data written by kernel task is 205GB over almost 20 hours of uptime. Which is much better than what I was seeing before. the system also seems less laggy than it was before.

Is it necessary to disable completely or can you just add the personal folders to Privacy?

 

[retta283]

Is it necessary to disable completely or can you just add the personal folders to Privacy?

You can add any folders to it. I'd add any folders that don't contain user documents or useful files, no need to see application files that never get opened manually. Personally I disabled it for my whole internal drive, as all my data is on an external.

 

[daverdfw]

You can add any folders to it. I'd add any folders that don't contain user documents or useful files, no need to see application files that never get opened manually. Personally I disabled it for my whole internal drive, as all my data is on an external.

I tried just only using it for applications, but the MDS process kept running, so I just added the entire drive to test my theory. So far I feel its working. But I will need more time to collect more data. I'd be curious for other people to try it and see.

 

[complaxneoda]

why m1 machine always show:

Available Spare Threshold: 99%

I checked my intel machine, this number is 2%

 

[saber106]

so just to report back after disabling spotlight things seem better, data written by kernel task is 205GB over almost 20 hours of uptime. Which is much better than what I was seeing before. the system also seems less laggy than it was before.

Can you take a screenshot of your spotlight option?

P/S: Guys, my iStatsMenu Status Written 1GB in 24hours of uptime, is that normal?

 

[complaxneoda]

Can you take a screenshot of your spotlight option?

P/S: Guys, my iStatsMenu Status Written 1GB in 24hours of uptime, is that normal?

may I ask how to get the written size of single application? thank you.

 

[saber106]

may I ask how to get the written size of single application? thank you.

Please turn on the Activity Monitor, goes to this tab and watch it

 

[daverdfw]

Can you take a screenshot of your spotlight option?

P/S: Guys, my iStatsMenu Status Written 1GB in 24hours of uptime, is that normal?

I am actually rebuilding my spotlight index now to test, in the exclusions you just hit the + symbol and add your entire drive

 

[patsfan83]

My 16GB 2TB M1 Mini. I have my iCloud library sync'd, at one point I copied over a 250GB aperture library and was importing it in to Lightroom, ended up stopping and deleting it all. Running 11.2 (was on Beta) and firmware is 1161.80.

Light browsing, mostly used to VPN and screen share in to my edit suite. Did that for 8 hours today and it read 10GB and wrote 4.5GB.


=== DRIVE HEALTH INDICATORS ===
ID | NAME | TYPE | UPDATE | RAW VALUE | VALUE | THRESHOLD | WORST | LAST MODIFIED | STATUS
2 Composite Temperature Life-span online 30 (30 °C) 70 30 70 2/24/21 7:48 PM 57.1% OK
3 Available Spare Pre-fail online 100 100 99 100 2/24/21 7:48 PM 100% OK
5 Life Percentage Used Life-span online 0 100 0 100 2/24/21 7:48 PM 100% OK
6 Data Units Read Life-span online 14,915,292 (6.9 TB) 100 0 100 2/24/21 7:48 PM 100% OK
7 Data Units Written Life-span online 14,151,587 (6.6 TB) 100 0 100 2/24/21 7:48 PM 100% OK
8 Host Read Commands Life-span online 82,470,337 100 0 100 2/24/21 7:48 PM 100% OK
9 Host Write Commands Life-span online 58,307,741 100 0 100 2/24/21 7:48 PM 100% OK
10 Controller Busy Time Life-span online 0 (0 minutes) 100 0 100 2/24/21 7:48 PM 100% OK
11 Power Cycles Life-span online 111 100 0 100 2/24/21 7:48 PM 100% OK
12 Power On Hours Life-span online 48 (48 hours) 100 0 100 2/24/21 7:48 PM 100% OK
13 Unsafe Shutdowns Life-span online 11 100 0 100 2/24/21 7:48 PM 100% OK
14 Media and Data Integrity Errors Pre-fail online 0 100 0 100 2/24/21 7:48 PM 100% OK
15 Error Information Log Entries Pre-fail online 0 100 0 100 2/24/21 7:48 PM 100% OK
16 Warning Composite Temperature Time Life-span online 0 (0 minutes) 100 0 100 2/24/21 7:48 PM 100% OK
17 Critical Composite Temperature Time Life-span online 0 (0 minutes) 100 0 100 2/24/21 7:48 PM 100% OK

 

[phoenix-mac-user]

Have had my Mac Mini since Mid-November and have used 4% so far. Which seems bad, but when I multiply that out should mean it will work for over 6 years, right?

Hoping Apple does something to fix it but if they don't I am fine with a computer I spent $700 on lasting 6 years.

Being totally honest, it will be on OfferUp in 3 years. lol

 

[patsfan83]

Have had my Mac Mini since Mid-November and have used 4% so far. Which seems bad, but when I multiply that out should mean it will work for over 6 years, right?

Hoping Apple does something to fix it but if they don't I am fine with a computer I spent $700 on lasting 6 years.

Being totally honest, it will be on OfferUp in 3 years. lol

My 2011 iMac SSD did .25% per month over 10 years or 19% as a comparison. Still runs great.

 

[Maximara]

However, this still highlights a couple of home truths:

1. Soldered-in SSDs - one of the components most likely to fail - are a bad idea. Full stop. An SSD is a consumable item with a limited number of writes, so it should be replaceable. I don't think there's anything good about a system which will predictably be "beyond economical repair" within 10 years... and even if the current scare is false, it is still perfectly feasible that an OS bug, a rogue app or just very heavy swap use could greatly reduce that lifespan. The current lifespan of SSDs is "acceptable" but there's not sufficient headroom to justify making them non-replaceable.

How does this prevent one from putting everything on an external SSD and booting from that bypassing this whole Soldered-in SSDs issue?

I have been booting from external drives since the 68k days and external SSDs aren't the money pit they once were. US$140 for 1TB via a big box store like Best Buy and that is more than $50 cheaper then the 1TB option from Apple and a 2TB is US$250 which is far less then the $600 Apple wants.

 

[thedocbwarren]

How does this prevent one from putting everything on an external SSD and booting from that bypassing this whole Soldered-in SSDs issue?

I have been booting from external drives since the 68k days and external SSDs aren't the money pit they once were. US$140 for 1TB via a big box store like Best Buy and that is more than $50 cheaper then the 1TB option from Apple and a 2TB is US$250 which is far less then the $600 Apple wants.

Why do you want to run a slower SSD externally when you have a super fast internal that doesn't require a dongle?

I'm unconvinced these reports are showing what people are concerned about or certainly not to the degree.

 

[flopticalcube]

T3 externals are fairly speedy but I wouldn't want a dongle hanging out of a laptop all the time if I could avoid it. Besides, it uses one of the only two available ports which is simply unacceptable.

 

[thedocbwarren]

T3 externals are fairly speedy but I wouldn't want a dongle hanging out of a laptop all the time if I could avoid it. Besides, it uses one of the only two available ports which is simply unacceptable.

AGreed, I don't see the reason to do that unless the internal SSD is extremely low capacity.

 

[osplo]

Stupid question, but, it seems that SMART is DISABLED on my disk. So, no "Percentage used" shown using smartmontools:

SMART support is: Available - device has SMART capability.
SMART support is: Disabled


Apparently I can enable it by running:

smartctl --smart=on --saveauto=on /dev/rdisk0 (Enables SMART on first disk)

but shouldn't I disable it afterwards? Or not? If I change the setting to smart=on might I see any adverse side effects (more data collected, slower SSD or something)?

Thanks.

 

[osplo]

Stupid question, but, it seems that SMART is DISABLED on my disk. So, no "Percentage used" shown using smartmontools:

SMART support is: Available - device has SMART capability.
SMART support is: Disabled


Apparently I can enable it by running:

smartctl --smart=on --saveauto=on /dev/rdisk0 (Enables SMART on first disk)

but shouldn't I disable it afterwards? Or not? If I change the setting to smart=on might I see any adverse side effects (more data collected, slower SSD or something)?

Thanks.

Well, finally I got this from my iMac SSD (to compare the writes in my good old Late 2012 disk) using sudo /usr/local/sbin/smartctl -a /dev/disk1s1:

SMART Attributes Data Structure revision number: 1
Vendor Specific SMART Attributes with Thresholds:
ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE
1 Raw_Read_Error_Rate 0x001a 200 200 000 Old_age Always - 0
5 Reallocated_Sector_Ct 0x0033 100 100 000 Pre-fail Always - 0
9 Power_On_Hours 0x0032 095 095 000 Old_age Always - 23885
12 Power_Cycle_Count 0x0032 094 094 000 Old_age Always - 5138
169 Unknown_Apple_Attrib 0x0013 253 253 010 Pre-fail Always - 3755110058752
173 Wear_Leveling_Count 0x0032 191 191 100 Old_age Always - 240535994521
192 Power-Off_Retract_Count 0x0012 099 099 000 Old_age Always - 136
194 Temperature_Celsius 0x0022 065 023 000 Old_age Always - 35 (Min/Max 12/78)
197 Current_Pending_Sector 0x0022 100 100 000 Old_age Always - 0
199 UDMA_CRC_Error_Count 0x001a 200 200 000 Old_age Always - 0

No way to see the bytes written.

Does anybody know how to interpret these figures? Anything out of bounds?

Thanks

 

[tab0reqq]

So definitely, you guys are not the only ones with such a high SWAP on SSD usage.

I bought my MBA M1 16/256 around the 5th of February.
As of today (26th February), I have over 9.4 TB written.

I think that's crazy.

Mine mainly use case:
3 profiles of Chrome (in summary 20-40 tabs, Gmail, FB Ads, Google Ads, internet marketing daily use case)
1 Safari with 10 tabs
1 Edge Chromium with 2-8 tabs opened
Spotify
1Password
iCloud
WhatsApp
Slack
Affinity Designer from time to time (editing logo/ads or similar)
OneNote
Mail
BBEdit
Reminders

Before M1, I've been doing the same ****, but much much less comfortable on a damn MBP 13 2015 with 8GB RAM.

My current MBA M1 16GB feels snappy as hell (compared to my prev. Asus G14 4800H/16GB/RTX2060), but that high SWAP on an SSD usage bogs me down.

 

[wirtandi]

I have been wanting to buy the MBA M1 for so long. My plan is to just go ahead and buy it, then run DriveDx t o see how much is written and read. If the amount is reasonable, good. If not, return. This plan sounds ok?