ssd swap - high usage of Terabytes Written

[Queen6]

Probably hypocritical of me to say this because with another head on I've been crowing loud about the M2 SSD-gate "scandal" but I think you're right in this instance. Stop looking at numbers and just use the bloody computer for what you bought it for.

I just think it's overblown. Apple's burnt itself on multiple occasions with the Mac, can you imagine the fallout of every single Mac being shown to be deliberately short lived due to the SSD...

I use my 2011 15" as an example, yet its not really a good example as is vastly differing technology. I'm of the mind that I don't care, nor will I unless the M1 SSD dies which I seriously doubt. TBH the weakest link in the battery and the users willingness to replace it...

Q-6

 

[Maximara]

One question I have: can you put the swap file on an external drive? It wouldn't be as fast as the internal storage but you wouldn't have the contention that you would normally get on the system when you are swapping. I was reading up on the Crucial MX 4 TB SSD and read that that model has 4 GB of DRAM cache so using it for swap could be very fast assuming that there's a lot of bandwidth to get to it.

AFAIK you can't put the swap file on an external drive. However, you can put applications and documents on the external as F-Train did (this will extend the life of the SSD as long as is possible):

 

[trip1ex]

My 2020 M1 Mac mini hasn't blown up yet you fear mongers you.

 

[topcat001]

I have 32 GB RAM on my 16" and my usual workflow/usage never touches swap. The Mac is basically a secondary device for me as all my machines run Linux (and OpenBSD) for both work and home usage. So far, the minimum amount written/day I have gotten is about 3.5 GB to 4 GB. Much of the Apple system is completely opaque so can't be exactly sure, but this is way too much for a mostly idle machine.

I know for my SSD this is not a concern but still it bothers me because in contrast my heavily used Linux systems might write maybe a 100 MB tops (usually much less) per day to the nvme (the Linux instrumentation is completely transparent so this is accurate). I would really like an option to disable some of this disk activity on macOS.

 

[Fishrrman]

topcat:
"I would really like an option to disable some of this disk activity on macOS."

Once again I'm going to mention this.
Have you tried TURNING OFF virtual memory disk swapping using the terminal?

 

[stigman]

I'm running now on Ventura Beta 5 and I have to say that data written has decreased in my case.

 

[philippe2014]

Hi, good news, so a problem has been solved at Apple.

 

[gank41]

My 2020 M1 MBP is showing mine as being 3% used. And I've been using it a LOT over the last (almost) two years...



Another note here- The 128 'Unsafe Shutdowns' were mostly during the early Monterey beta cycle. Probably ½ of them all occurred in a month period or so... And the rest from when I was on Big Sur initially. I've had less and less of those lately.

 

[daavee80]

My 2020 M1 MBP is showing mine as being 3% used. And I've been using it a LOT over the last (almost) two years...

View attachment 2047775

Another note here- The 128 'Unsafe Shutdowns' were mostly during the early Monterey beta cycle. Probably ½ of them all occurred in a month period or so... And the rest from when I was on Big Sur initially. I've had less and less of those lately.

Can I ask what size SSD this is? That percentage used indicates that the NAND will be good for aroiund 5 Petabytes of writes. Which is an enormous amount if this is standard consumer grade NAND.

Either Apple is being very generous with their TBW estimates or they're using enterpirse grade NAND for these consumer machines. I think it's obvious which one of those is the more realistic notion...

I will read through the whole thread at some point as I'm sure this has been discussed already but yeah, I'm just surprised that any one of these 'drives' is rated at 5000 TBW.

 

[gank41]

Can I ask what size SSD this is? That percentage used indicates that the NAND will be good for aroiund 5 Petabytes of writes. Which is an enormous amount if this is standard consumer grade NAND.

Either Apple is being very generous with their TBW estimates or they're using enterpirse grade NAND for these consumer machines. I think it's obvious which one of those is the more realistic notion...

I will read through the whole thread at some point as I'm sure this has been discussed already but yeah, I'm just surprised that any one of these 'drives' is rated at 5000 TBW.

This is the 2020 M1 MBP with the 2TB/16GB configuration.

 

[daavee80]

This is the 2020 M1 MBP with the 2TB/16GB configuration.

Ah, ok thanks.

 

[jdb8167]

Can I ask what size SSD this is? That percentage used indicates that the NAND will be good for aroiund 5 Petabytes of writes. Which is an enormous amount if this is standard consumer grade NAND.

Either Apple is being very generous with their TBW estimates or they're using enterpirse grade NAND for these consumer machines. I think it's obvious which one of those is the more realistic notion...

I will read through the whole thread at some point as I'm sure this has been discussed already but yeah, I'm just surprised that any one of these 'drives' is rated at 5000 TBW.

One point is that Apple can afford to be as accurate as possible since they aren’t using the TBW value for warranties. This is different from a commercial SSD vendor.

 

[daavee80]

One point is that Apple can afford to be as accurate as possible since they aren’t using the TBW value for warranties. This is different from a commercial SSD vendor.

You are of course right. It has been proven that at least some SSD vendors have been known to underestimate the TBW values oif their products..
Maybe I was a bit hasty in stating that Apple are likely over-estimating the endurance of these drives. There are loads of things to look in to - their supply chain for a start - and besides I'm no engineer. Like I said I was just surprised by the lifetime estimation. Looking back through the thread it seems that the 256GB drives for example are estimated for 600 TBW so 5000 for the 2TB fits that pattern.

Although I have to say it would seem a bit of a leap if turns out they're using the same NAND that is sold in off-the-shelf consumer drives with warranty ratings of around 400-600 TBW per Terabyte. To go from that to four times that amount seems rather incredible given that SSD vendors would surely jump at the chance to market their SSD's as having, say, twice the warrantied endurance while still being able to play it massively safe by undeclaring the actual TBW value by 50%!

 

[jdb8167]

You are of course right. It has been proven that at least some SSD vendors have been known to underestimate the TBW values oif their products..
Maybe I was a bit hasty in stating that Apple are likely over-estimating the endurance of these drives. There are loads of things to look in to - their supply chain for a start - and besides I'm no engineer. Like I said I was just surprised by the lifetime estimation. Looking back through the thread it seems that the 256GB drives for example are estimated for 600 TBW so 5000 for the 2TB fits that pattern.

Although I have to say it would seem a bit of a leap if turns out they're using the same NAND that is sold in off-the-shelf consumer drives with warranty ratings of around 400-600 TBW per Terabyte. To go from that to four times that amount seems rather incredible given that SSD vendors would surely jump at the chance to market their SSD's as having, say, twice the warrantied endurance while still being able to play it massively safe by undeclaring the actual TBW value by 50%!

People have tested SSDs to failure in the past and always got way beyond the TBW rating. Sometimes by many multiples.

 

[evertjr]

You are of course right. It has been proven that at least some SSD vendors have been known to underestimate the TBW values oif their products..
Maybe I was a bit hasty in stating that Apple are likely over-estimating the endurance of these drives. There are loads of things to look in to - their supply chain for a start - and besides I'm no engineer. Like I said I was just surprised by the lifetime estimation. Looking back through the thread it seems that the 256GB drives for example are estimated for 600 TBW so 5000 for the 2TB fits that pattern.

Although I have to say it would seem a bit of a leap if turns out they're using the same NAND that is sold in off-the-shelf consumer drives with warranty ratings of around 400-600 TBW per Terabyte. To go from that to four times that amount seems rather incredible given that SSD vendors would surely jump at the chance to market their SSD's as having, say, twice the warrantied endurance while still being able to play it massively safe by undeclaring the actual TBW value by 50%!

You cannot compare the new SSDs on AS Macs to others SSDs on the market since these were designed for Apple within they specified specs, im pretty sure they are overprovisioned as the SSD now play a key role in the unified memory architecture. I don't know where did you get the idea the 256 drivers are rated at 600TBW since there's people here way beyond that as result of memory leaks that was only fixed recently (https://forums.macrumors.com/threads/ssd-swap-high-usage-of-terabytes-written.2284893/post-30992637) but still within the reported life percentage range... My 256 drive is at 105TBW and only at 7% percentage used, so Apple expects it to last around 1.7 Petabytes and I have no doubts it will.

 

[Maximara]

You cannot compare the new SSDs on AS Macs to others SSDs on the market since these were designed for Apple within they specified specs, im pretty sure they are overprovisioned as the SSD now play a key role in the unified memory architecture. I don't know where did you get the idea the 256 drivers are rated at 600TBW since there's people here way beyond that as result of memory leaks that was only fixed recently (https://forums.macrumors.com/threads/ssd-swap-high-usage-of-terabytes-written.2284893/post-30992637) but still within the reported life percentage range... My 256 drive is at 105TBW and only at 7% percentage used, so Apple expects it to last around 1.7 Petabytes and I have no doubts it will.

Yes and this is what irked me with this whole thing — everybody went crazy over the TBW amounts but rarely looked at the percentage used figure. The moment you looked at that (if it was even given to begin with) the majority of 'ah the MacOS is overwriting to the SSD' thing turned into a big Nothing Burger with a side order of Nothing Fries and a Nothing Drink.

In the few cases where something clearly was wrong it could be teased out that is was some other program (Chrome or MS Team, anything made with Electron, some Intel programs) causing the write issues rather than anything actually endemic to the MacOS itself. I will blame Apple for messing up Safari to the point I went back to Firefox.

 

[daavee80]

You cannot compare the new SSDs on AS Macs to others SSDs on the market since these were designed for Apple within they specified specs, im pretty sure they are overprovisioned as the SSD now play a key role in the unified memory architecture. I don't know where did you get the idea the 256 drivers are rated at 600TBW since there's people here way beyond that as result of memory leaks that was only fixed recently (https://forums.macrumors.com/threads/ssd-swap-high-usage-of-terabytes-written.2284893/post-30992637) but still within the reported life percentage range... My 256 drive is at 105TBW and only at 7% percentage used, so Apple expects it to last around 1.7 Petabytes and I have no doubts it will.

I hadn't gotten around to reading the entire thread and so hadnt seen these more recent posts, which is why I was working with a rough figure of 600TBW for 256GB. So yeah, it appears that this is some decently high-endurance NAND Apple are using. And you're right, Apple's reliance on the SSD for swap would require such NAND so it does make sense.

Yes and this is what irked me with this whole thing — everybody went crazy over the TBW amounts but rarely looked at the percentage used figure. The moment you looked at that (if it was even given to begin with) the majority of 'ah the MacOS is overwriting to the SSD' thing turned into a big Nothing Burger with a side order of Nothing Fries and a Nothing Drink.

In the few cases where something clearly was wrong it could be teased out that is was some other program (Chrome or MS Team, anything made with Electron, some Intel programs) causing the write issues rather than anything actually endemic to the MacOS itself. I will blame Apple for messing up Safari to the point I went back to Firefox.

The TBW figure is what SSD vendors and manufacturers use and the software-obtained percentage figure just didn't seem feasible to a lot of people, me included. So you can see where the confusion and room for interpretation of the figures derived from.

People have tested SSDs to failure in the past and always got way beyond the TBW rating. Sometimes by many multiples.

Of course. And given that and all of the above I think that either - (A) every single one of these vendors is somehow missing a trick in underselling the endurance of their SSD's or (B) Apple is using relatively expensive enterprise grade NAND or (C) Apple know something we don't.
Hmm, either way I'm not sure I'm too worried any more. If there are 256GB drives in the wild still working perfectly with well over a Petabyte written to them then Apple could be buying the chips from Satan himself for all I care, as long as my Macbook SSD can stand up to many years of heavy writing.

 

[Maximara]

The TBW figure is what SSD vendors and manufacturers use and the software-obtained percentage figure just didn't seem feasible to a lot of people, me included. So you can see where the confusion and room for interpretation of the figures derived from.

The thing is that TBW figure is what SSD vendors and manufacturers use is what they will insure for

"However, not all published TBW ratings are pure—or even accurate—estimates of a drive’s lifespan. Most are the result of a complicated financial formula that of course includes expected lifespan, but also factors in pricing plus the cost of support and replacement. Why?

If a drive fails within the warranty period before its TBW rating is reached, according to most warranties, the vendor has to replace it. If it fails after surpassing the TBW rating, the vendor does not have to replace it, even if the warranty period has not lapsed.
(...)
Lower TBW rating keeps prosumers and business users from buying cheaper client drives, working them to death, and then expecting them to be replaced for free. (...) Basically, the TBW rating/warranty deal is a game rigged slightly in the house’s favor." — TBW: How this obscure rating can affect your SSD’s warranty

 

[osplo]

My 2020 M1 MBP is showing mine as being 3% used. And I've been using it a LOT over the last (almost) two years...

View attachment 2047775

That´s great. Your disk should last for another 64 years of heavy usage.

 

[Maximara]

Of course. And given that and all of the above I think that either - (A) every single one of these vendors is somehow missing a trick in underselling the endurance of their SSD's or (B) Apple is using relatively expensive enterprise grade NAND or (C) Apple know something we don't.
Hmm, either way I'm not sure I'm too worried any more. If there are 256GB drives in the wild still working perfectly with well over a Petabyte written to them then Apple could be buying the chips from Satan himself for all I care, as long as my Macbook SSD can stand up to many years of heavy writing.

The SSD Endurance Experiment: They’re all dead article (and Youtube video) from 2015 shows (A) is the most likely option.

The Samsung’s 840 Pro 256GB lasted 2.4 PB (2,400 TB) before finally dying. Samsung only warrents the 840 Pro 256GB drive for 10 years or 150 TB TBW. A real world test showed the 840 Pro 256G drive lasting 16x what it was warrented far in terms of TBW. Heck, the first hiccup in testing that drive didn't appeared until 200 TB well past the 150 TB mark and only failed a hash check at 300TB.

 

[DaKKs]

My 2021-March M1 Air 8/256 has, according to dxdrive, 28.0TB written to it. 566 hours of use. Reported as 98% health.

 

[shimpster]

Finally got around to getting this on my M1 MacBook Pro 2tb/16GBRam Option purchased back on April 5th 2021. Really not sure where they 28 unsafe shutdowns came from. But I think this is pretty good?

 

[osplo]

Finally got around to getting this on my M1 MacBook Pro 2tb/16GBRam Option purchased back on April 5th 2021. Really not sure where they 28 unsafe shutdowns came from. But I think this is pretty good?
View attachment 2064752

Yeah, 1% used in more than one year and five months means you are good for 99 years more.

Unless you are going to be really long-lived you are on the safe side, I'd say.

 

[theorist9]

You are in the wrong thread as this is for Apple Silicon computers. However if your 2019 Intel computer is relatively new as you mentioned (!) chances are that the initial Data Units Written was high right from the start (OS installs, factory checks, apps installs and all kind of initialization). My two cents: check again in a month and see the delta to see a reliable measure. Install 12.5 first, it might help.

The 2019 iMac is not new, but the SSD is. It was indeed installed 59 days before I posted.

You are of course right that this is an AS thread, but I though offering an Intel Mac running a current OS as a point of comparison would be useful.

Your guess about initialization, while plausible and good idea to check, turned out to be unfounded. I just checked again, 162 days later. Data written to my iMac's new 2TB WD SN850 has increased from 8.71 TB to 32.5 TB, so we have:

(32.5 - 8.71)/162*1000 = 147 GB/day, which is essentially unchanged from the 148 GB/day I found for the first two months of SSD use. I.e., the high writes have nothing to do with initialization.

Fortunately this is not a concern, since my drive's endurance is 1200 TBW which, at 150 GB/day, will not be reached for 22 years.

I've kept Monterey up to date, and am currently on 12.6.2 (can't switch to Ventura yet because my university's security systems aren't ready).

 

[BENCHPRESS]

M1 Max 1TB 32GB

Device currently has an uptime of 6 days and it's written 735GB.

kernel_task = 250GB
photoanalysisd = 220GB
launchd = 83GB
corespotlightd = 65GB

are the biggest culprits. I don't do anything but browse the web. No development or production work. Pure web browsing and Spotify.

Anyone know why these processes are writing so much to the disk? I'm not even doing much with the device.. Already disable virtual memory swap, limited spotlight indexing to only important folders. Why the hell is photoanalysisd writing so much anyway?

Any help appreciated.