Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
Data Storage IT Technology

Apple M1 Mac Users Report Excessive SSD Wear (macrumors.com) 274

Over the past week, some M1 Mac users have been reporting alarming SSD health readings, suggesting that these devices are writing extraordinary amounts of data to their drives. From a report: Across Twitter and the MacRumors forums, users are reporting that M1 Macs are experiencing extremely high drive writes over a short space of time. In what appear to be the most severe cases, M1 Macs are said to be consuming as much as 10 to 13 percent of the maximum warrantable total bytes written (TBW) value of its SSD. Flash memory on solid-state drives, such as those used in Macs, can only be written to a certain number of times before they become unstable. Software ensures that load is spread evenly across the drive's memory cells, but there is a point when the drive has been written to so many times that it can no longer reliably hold data. So while SSD wear is normal, expected behavior, drives should not be exhausting their ability to hold data as quickly as some M1 Macs seem to be. One user showed that their M1 Mac had already consumed one percent of its SSD after just two months, while another M1 Mac with a 2TB SSD had already consumed three percent. The total data units written for these machines is running into many terabytes, when they would normally be expected to be considerably lower.
This discussion has been archived. No new comments can be posted.

Apple M1 Mac Users Report Excessive SSD Wear

Comments Filter:
  • First gen products (Score:5, Insightful)

    by slazzy ( 864185 ) on Tuesday February 23, 2021 @09:39PM (#61094392) Homepage Journal
    That's why I never buy first gen products, there's always problems. Hopefully this turns out to be a software fix, not a "you're holding it wrong" situation.
    • by dgatwood ( 11270 ) on Tuesday February 23, 2021 @10:05PM (#61094430) Homepage Journal

      That's why I never buy first gen products, there's always problems. Hopefully this turns out to be a software fix, not a "you're holding it wrong" situation.

      I'd imagine it's almost certainly a "you're holding it wrong" situation.

      The main problem is that all of these machines have too little RAM for a device that has to support two different CPU architectures AND is flash-based. Apple arguably shouldn't even sell computers with only 8 GB of RAM these days, but they absolutely should not be selling machines running a new CPU architecture with such inadequate RAM.

      With only 8 GB of RAM, these machines are likely to be constantly under memory pressure because of trying to hold enough of both native AND translated Intel versions of various frameworks in RAM. That is likely to result in a dramatically higher rate of paging than on an Intel-architecture-only device and, barring a miracle, will also result in a dramatically higher rate of paging out dirty pages to disk. And boom. Now you have a computer burning through one eighth of its flash life within the first three months.

      But whether that theory is correct or not, for people who have already lost an eighth of their device's usable life in only three months, it kind of doesn't matter what the root cause is at this point; they've burned through a couple of years of flash life already, and there's a decent chance that their hardware won't make it through the AppleCare warranty period. Those folks need to back up daily, buy AppleCare, and expect to end up getting a new motherboard in two years. And if that sort of burn rate is common, these machines could be a very expensive design mistake for Apple.

      • by dgatwood ( 11270 ) on Tuesday February 23, 2021 @10:16PM (#61094456) Homepage Journal

        But whether that theory is correct or not, for people who have already lost an eighth of their device's usable life in only three months, it kind of doesn't matter what the root cause is at this point; they've burned through a couple of years of flash life already, and there's a decent chance that their hardware won't make it through the AppleCare warranty period. Those folks need to back up daily, buy AppleCare, and expect to end up getting a new motherboard in two years. And if that sort of burn rate is common, these machines could be a very expensive design mistake for Apple.

        I think I understated things a bit. I just looked at the article. One user reported 150 TB of writes. My heavily used 2017 MBP is still under 30 TB of writes after five years. That means that the user in question basically saw the equivalent about one decade of heavy power user use every month. If this isn't a software reporting bug, that machine is toast.

        • by gnasher719 ( 869701 ) on Wednesday February 24, 2021 @04:21AM (#61094958)
          150 TB of writes in 2 months. Thats 2,500 GB a day. That's 29 MB per second, every second of the day, for two months uninterrupted.

          What is more likely: That your Mac is writing 29 MB of data every second, or that some tool that is supposed to record the number of writes gets it completely wrong? I'd like to know if these guys bothered turning on Activity Monitor and see what happens.
          • by dgatwood ( 11270 ) on Wednesday February 24, 2021 @08:35AM (#61095294) Homepage Journal

            150 TB of writes in 2 months. Thats 2,500 GB a day. That's 29 MB per second, every second of the day, for two months uninterrupted.

            What is more likely: That your Mac is writing 29 MB of data every second, or that some tool that is supposed to record the number of writes gets it completely wrong? I'd like to know if these guys bothered turning on Activity Monitor and see what happens.

            It's the hardware that records the number of writes, and that's happening down at the firmware level. A bug in something that simple (if (isWrite) writtenBlocks++;) seems kind of unlikely to me. I think it's more likely that some early hardware shipped with a non-zeroed statistics page or that some small percentage of hardware got an obscene amount of stress testing at the factory.

            • by guruevi ( 827432 ) on Wednesday February 24, 2021 @09:39AM (#61095442)

              The SMART status of disk drives is notoriously hard to interpret. If I read the raw value of temperature on Seagate devices and interpreted it literally, they would be on fire while some SSD would be in liquid nitrogen.

              It could easily be a buffer overflow or the SMART tool not having the latest vendor lookup table. I've noticed that some SSD's report their wear levels 'backwards' so they'll report 98% wear (which should be interpreted as 2% used). Then there are a host of other vendors (Samsung especially) that simply report everything useful in a proprietary format.

              • The SMART status of disk drives is notoriously hard to interpret.

                It really isn't for SSD values. The raw value is always given in LBAs written. If you read that literally without conversion it would on a typical SSD be around 512x smaller than in reality, and not grossly overreported.

                There is no lookup table required here. Unlike many of the SMART values from spinning rust days what was adopted for SSDs (smart value 241) is standard across manufacturers and brands.

        • by AmiMoJo ( 196126 ) on Wednesday February 24, 2021 @05:52AM (#61095088) Homepage Journal

          It's okay though, Apple can just fix the bug, swap the SSD out and copy the user's data over.

          Oh wait, the SSD is soldered to the motherboard, along with everything else. I guess they will have to do an extended warranty and replace the entire motherboard when it dies.

          • by The123king ( 2395060 ) on Wednesday February 24, 2021 @06:33AM (#61095120)
            This sort of fuckup could cost Apple dearly, if they are made to recall, or at least replace all faulty motherboards. The upside is, since they might get bitten by this in the wallet department, they may see replaceable SSD's as a cheaper way to mitigate such an expensive issue in the future.
            • by AmiMoJo ( 196126 ) on Wednesday February 24, 2021 @06:58AM (#61095144) Homepage Journal

              Apple's usual policy is to deny the problem, then claim there is a software fix, and eventually a few years and a couple of lawsuits down the line do an extended warranty. The goal is to make sure most of the affected products are in landfill before they accept any liability, and then keep it as quiet as possible.

              • by kurkosdr ( 2378710 ) on Wednesday February 24, 2021 @08:34AM (#61095290)
                My spidey sense tells me that this time they will declare the SSD to be a "consumable item" (consumable items have no warranty or a warranty measured in single-digit months), despite the "consumable item" in question being soldered-on. They are already doing it with their non-replaceable batteries.
            • by thereddaikon ( 5795246 ) on Wednesday February 24, 2021 @08:55AM (#61095342)

              This sort of fuckup could cost Apple dearly, if they are made to recall, or at least replace all faulty motherboards. The upside is, since they might get bitten by this in the wallet department, they may see replaceable SSD's as a cheaper way to mitigate such an expensive issue in the future.

              hahaha no. Apple is one of the most valuable companies in the world if not the most. They are going to continue with hyper integration and anti consumer practices because for every debacle like this its more than offset by the increased sales they farm from forced obsolescence.

              The only way this stops is if the law changes. Apple will not willfully do so. They have no reason to do so. Apple, like all companies, doesn't care about you or the environment or anything beyond making more money.

      • ikely to be constantly under memory pressure because of trying to hold enough of both native AND translated Intel versions of various frameworks in RAM

        I don't see how that makes sense.

        The native ARM libraries would tend to be shared across multiple applications.

        Rosetta translated intel apps would have the application code translated, but would still be calling into native system libraries, still shared. Other third party libraries they had (like physics engines) would be translated, yes, but that's just si

        • by dgatwood ( 11270 )

          ikely to be constantly under memory pressure because of trying to hold enough of both native AND translated Intel versions of various frameworks in RAM

          I don't see how that makes sense.

          The native ARM libraries would tend to be shared across multiple applications.

          Rosetta translated intel apps would have the application code translated, but would still be calling into native system libraries, still shared.

          Only if they're doing things very differently this time around than they did in the first Rosetta. With the original Rosetta, the translation occurred at the system call level, i.e. every single framework was a multi-architecture binary, and the PowerPC versions of apps ran the PowerPC versions of libraries and frameworks in emulation mode.

          • by Nebulo ( 29412 ) on Tuesday February 23, 2021 @11:09PM (#61094558)

            It is very different this time, actually. Intel binaries are translated to the ARM instruction set once, the first time they are launched (or loaded, if we're talking a library) and the results cached to disk. No Intel code is ever loaded into RAM for the purposes of execution.

            • by dgatwood ( 11270 )

              That doesn't really answer the question of whether they're running translated versions of the Intel frameworks or the true native ARM frameworks, though. The latter would be amazing, but would have a much higher risk of bugs if there are any architecture-specific differences in the behavior of literally any library or framework.

        • by 93 Escort Wagon ( 326346 ) on Wednesday February 24, 2021 @01:00AM (#61094720)

          I would wager it's some sort of MDS/Spotlight aberration, that thing can get pretty chatty hard-drive wise on a normal system...

          I was wondering about that as well, although the amount of read / written data being reported is pretty incredible. It's hard to imagine Spotlight thrashing the SSD that much.

          When I run smartctl on the brand-new M1 Air I got just a couple weeks ago (running Big Sur, obviously), it already reports reads / writes of 553 GB / 564 GB (with 0% used, thank goodness). Contrast that to the 2015 MacBook Pro I've had for several years now - it's running Mojave, and reports a relatively paltry 4.35 TB / 4.81 TB (and only 1% used).

          • When I run smartctl on the brand-new M1 Air I got just a couple weeks ago (running Big Sur, obviously), it already reports reads / writes of 553 GB / 564 GB (with 0% used, thank goodness).

            That may be a bit on the high side, but really not a problem. Someone reporting 150TB in two months is madness. (You'll reach 150 TB in about 260 times "couple weeks", so that would be many years).

            150TB in two months looks like someone installed a benchmark that tests read/write speeds and kept it running for two months. Or they used a measuring tool that is totally off.

      • âoethey absolutely should not be selling machines running a new CPU architecture with such inadequate RAM.â

        By they do you mean everyone? I, unfortunately, know for a fact that at least Dell and HP still sell mechanical HD, integrated graphics (meaning part of that paltry amount of RAM is forfeit) machines with 4GB of RAM running Windows 10. While they donâ(TM)t experience âexcessive SSD wearâ(TM) (what with not having a solid state drive) they do experience excessive running like ga

        • by CaptQuark ( 2706165 ) on Wednesday February 24, 2021 @01:57AM (#61094770)

          These are not really high end machines mind you, but they are very much in the general M1 price range. If you think I exaggerate, get a new i5 range laptop...

          I don't know where you are getting your pricing information, but I'll pull prices from the manufacturers websites.

          The Apple MacBook Pro: Apple M1, 8MB RAM, 256GB SSD $1299

          Dell Inspiron 15 3000: Intel i3, 4MB RAM, 128GB SSD $299
          Dell Inspiron 15 3000: Intel i3, 8MB RAM, 256GB SSD $379
          Dell Inspiron 15 3000: Intel i3, 8MB RAM, 1TB HD $449 (Only non-SSD Dell offers)
          Dell Inspiron 15 3000: Intel i5, 8MB RAM, 256GB SSD $499
          Dell Inspiron 15 3000: AMD Ryzen 5, 8MB RAM, 512GB SSD $499
          Dell Inspiron 15 3000: Intel i7, 12MB RAM, 512GB SSD $649
          Dell Inspiron 15 5000: Intel i7, 8MB RAM, 256GB SSD $649
          Dell Inspiron 15 5000: Intel i7, 12MB RAM, 512GB SSD $749
          Dell XPS 13: Intel i7, 16MB RAM, 512GB SSD $879

          The Dell i7 highlighted has twice the storage, 50% more RAM, and is still half the price. I don't consider that a hot, power consuming paperweight either.

          --

      • by SynKKnyS ( 534257 ) on Wednesday February 24, 2021 @02:22AM (#61094800)
        I think this is it. I have a 16GB ram model and my wear is nowhere close (about 1TB) in the same 2 month timespan. I don't use my device lightly either as I always have a mix of VS Code, Lightroom/Photoshop, AdGuard, Blender, etc. open.
        • Also, I'd imagine a good amount of those writes are from FileVault and iCloud doing their first time encryption/syncing when I setup the device.
      • Why would the m1 load both the entire arm and x86 versions of a binary ?
      • If they borrowed/recompiled zram, and made clever use of tmpfs like file stores in their file system, this could be strongly mitigated.

        zstd compression algo on zram swap is often able to get between 2x and 3x compression on normal data. Putting swap on an SSD is always stupid, no matter what.

        I use similar mitigation processes on my hacked chromebook.

      • With only 8 GB of RAM, these machines are likely to be constantly under memory pressure because of trying to hold enough of both native AND translated Intel versions of various frameworks in RAM.

        Intel apps and frameworks are translated ONCE and then they behave just alike ARM apps and frameworks. (Exception: Apple doesn't delete the Intel code, so if they released an improved translator, they would translate every app once more). But anyway, that would be at most a few gigabyte, not the numbers reported.

        • by dgatwood ( 11270 )

          But the point is that they presumably still use the translated Intel frameworks, rather the native ARM frameworks, when running a translated binary. Assuming that's true, then there are still potentially two separate copies of some of those framework binaries loaded into memory at any given point in time whenever you're running a mix of native and non-native apps.

      • The main problem is that all of these machines have too little RAM for a device that has to support two different CPU architectures AND is flash-based. Apple arguably shouldn't even sell computers with only 8 GB of RAM these days, but they absolutely should not be selling machines running a new CPU architecture with such inadequate RAM.

        You are absolutely correct. Especially with Big Slur - oops, Sur. Which I point my finger at.

        I have a 2019 Intel iMac that ran along pretty good on 8GB before moving to Big Sur, which is unfortunately kind of Apple's Windows 8. Because Big Sur is definitely not ready for prime time. After the update, every thing has gone into a time warp - keyboard lag, bluetooth problems, Video frame rate slowdown. Just taking a decent computer and strangling something. If I let it sit for a half hour after boot, it im

    • Even if there is a software fix to prevent it from happening going forward, it doesn't replace all of the write cycles used since then. I don't know what kind of NAND they're using, but assuming it's the newer QLC, the maximum write cycles is comparatively low at around a thousand or so as compared to the tens of thousands of cycles that MLC drives tended to have.

      I have a feeling there's going to end up being a lawsuit over this and Apple might wind up doing some drive replacements for free.
      • You mean solder in new drives?

        Fat chance.

        They will fight it to their death, almost, and then claim they always welcomed replacing units, and give you an elaborate process of burning hoops to jump through, to have the entire machime replaced with all your data lost and you paying pretty much the full price a second time anyway unless yoi got through all those hoops that in practice nobody will ever be able to do, but lets them claim you can, so you can't sue them again.

      • QLC is wholly inappropriate for most of the available sizes on the M1 gear, and the worst possible choice from a technical standpoint in all the default configurations (256GB.) since density doesnt have to be an issue. Four 64GB TLC chips are more than small enough already.
    • The problem is that people are running multi-app desktop workloads on a machine that has iPhone size RAM causing it to page out RAM continuously to the SSD due to too much stuff running on it.

      These workstation level users should have waited for a proper desktop version that has more RAM, but instead we are here now at this problem. I don't see Apple coming up with a software fix, unless they suspend background apps that are paged out to the SSD.

      • by tysonedwards ( 969693 ) on Tuesday February 23, 2021 @11:11PM (#61094564)
        Not really. Having personally looked at this issue and reported bugs on it during the DTK period, the issue is within firmware and one of an improper TRIM implementation that causes cells to be continually overwritten instead of when a block is dereferenced. One doesn’t even need booted into the OS for this behavior. Simply sitting at the boot menu or internet recovery menu is enough. But that requires Apple look at their bug reports, let alone try to reproduce the issue instead of closing the ticket with “Likely resolved following the release of macOS 11.0.1. If issue persists, please create a new Feedback.”
    • Hopefully this turns out to be a software fix, not a "you're holding it wrong" situation.

      What is there to fix??? The actual numbers they quote are pretty damn good. Three percent wear after two months is five and a half years life expectancy. That's the generally accepted useful life for an SSD in normal use. The other example, one percent after two months, is almost 17 years of life expectancy.

      In any case, the best thing to do for a computer with an SSD is to immediately buy an SSD which is identical to the one inside the computer. It's a great backup scheme, for one, to take an image of

      • That's the generally accepted useful life for an SSD in normal use

        No. At 5 years, my Samsung drive will just be going out of warranty. Hard drives have a 1 year warranty. The useful lifespan for average use is way higher for an SSD than an HDD.

        Good luck swapping out your soldered SSD any time you like...

  • by BoogieChile ( 517082 ) on Tuesday February 23, 2021 @09:42PM (#61094400)
    A fast /swap drive will do that for you.

    Of course, it'll do this too.
    • by dfghjk ( 711126 )

      That demonstrates too great an understanding of the underlying technology to be accepted here. Macs make better use of memory because of magic. 8GB is fine, where's my check?

  • by PhunkySchtuff ( 208108 ) <.kai. .at. .automatica.com.au.> on Tuesday February 23, 2021 @09:48PM (#61094410) Homepage

    One percent of total drive life used up after two months implies that the drive will last, under current usage patterns, for 200 months. That's over 16 years. I don't see the problem.
    Even three percent, after some unspecified time, isn't excessive. Let's assume it's 3% in 2 months, that's still 5.5 years of usage at current levels. Not stellar, but well in excess of Apple's 3 year warranty. No, it's not ideal that the computer will die in 5 1/2 years and be unrepairable due to the soldered flash chips.

    • by sodul ( 833177 )

      In some countries having the machine very frequently die after 5.5Y will probably be legally flagged as a hidden manufacturing defect, for which warranty does not expire.
      There is a hight risk that Apple will need to replace the SSD on Million of machines if they die this quickly. To be fair anything over 10y for electronics is a good long run, but 5.5Y for an Apple device is not great.

      • Agreed.

        Sent from my Mac Mini 2012.
      • by v1 ( 525388 )

        In some countries having the machine very frequently die after 5.5Y will probably be legally flagged as a hidden manufacturing defect,

        The "durability" of a product varies considerably depending on what sort of product it is. A car that wears out after 6 years, yeah that's a problem. But that's WAY longer than anyone expects their shoes to last. And that's longer than the majority of people keep their cell phone before upgrading. This tends to be tied to the price of a product, because people expect to g

    • A bit faster. You need to calculate in the acceleration of destruction as the usable portion of the drive effectively shrinks over time and takes increasingly faster hammering.

      • Also, if you can't use half the damn drive anymore after 2.75 years, that's a lawsuit!

        • Its called wear leveling not wear focusing

          Pretty much every time you post I cant understand why you made the choice to do so, and cannot fathom how someone demonstrating such a gross lack of any knowledge on the subject would pretend that they had anything keen to add at all.
  • by Rosco P. Coltrane ( 209368 ) on Tuesday February 23, 2021 @09:51PM (#61094412)

    One user showed that their M1 Mac had already consumed one percent of its SSD after just two months

    So the lifespan of the device is roughly 200 months without the user saving anything of course. Let's say the user user fills up their 90% of their hard drive: that's 10% free space to do wear levelling on, so still 20 months.

    But this is an Apple product: in 20 months, it won't be the latest and greatest shiny no more. Chances are it'll be resold and replaced long before that.

    • Re:Seems alright (Score:5, Informative)

      by ShanghaiBill ( 739463 ) on Tuesday February 23, 2021 @10:24PM (#61094474)

      Let's say the user user fills up their 90% of their hard drive: that's 10% free space to do wear levelling on, so still 20 months.

      SSDs don't work that way. The permanent data will be shifted around so the frequently written temporary data is spread over the entire drive.

      • Indeed. It seems a lot of posters here dont really understand the term wear leveling or what such an algorithm must do, even to the most static of files that only ever get written once in a decade.
  • by Proudrooster ( 580120 ) on Tuesday February 23, 2021 @10:09PM (#61094440) Homepage

    In the M1 SoC Architecture, the CPU and GPU share the same memory and with 8 measly jigabytes there is sure to be some swap to disk activity. (Swap just means less used stuff is copied to the SSD to make room for the new stuff going into memory). 8GB of memory might have been fine when the GPU had it's own memory buy now it is cutting into program memory.

    In my late 2013 Mac with the Radeon RX580, the memory card has it's own private 8GB of memory and the system has 16GB. That's a total of ummmm 20+ GB and the new M1 has only 30%. That is a sure sign that swapping is going to occur.

    The M1 architecture is fast and power efficient but the tradeoff is upgradability, you'd think Apple could at least stick an M.2 slot in there for more storage.

    So if you are going to buy one of this little beasties and use it for more that surfing the web, listening to music, and watching cat videos, you'd best go with the 16 GB model but you'd better save your pennies, going from 8GB with a 256GB SSD to 16GB with a 512GB SSD will cost you $500. Ouch!

    Considering a retail priced gaming 500GB SSD is $50 and 16GB of gaming DDD4 memory is $100.
    Apple must be charging for the cost to desolder the old stuff and resolder the new stuff.

    • Comment removed (Score:4, Insightful)

      by account_deleted ( 4530225 ) on Tuesday February 23, 2021 @10:39PM (#61094506)
      Comment removed based on user account deletion
    • Apple or somebody should do a memory trace and see what the cause is. Paging is not linear to memory use - it is a logarithmic curve, that ends in an undesirable state call thrashing. Normally the os scheduler knows this, and prevents any new tasks from starting, bar system dumps. The good thing about China, is for previous models, there are phone shops that specialise in ram and memory upgrades, close to the actual price of the memory. Hopefully M1 will be surface mount repairable/upgradable by component
    • The storage upgrade is just having USB ports. I've hung a few drives off of mine, so I've got gobs of storage, just very little of it is inside the computer. Given the speed and price of external storage these days—spinning drives for media and mass storage, external SSD drives for things that need to be fast—it's not as relevant to me to have a massive internal drive anymore.

      I've got the 16GB/256GB mini, and my numbers are not meaningfully different than the ones posted. They're slightly better

  • the hard drive is glue in.
  • Sensational. (Score:3, Insightful)

    by msauve ( 701917 ) on Tuesday February 23, 2021 @10:44PM (#61094518)
    "One user showed that their M1 Mac had already consumed one percent of its SSD after just two months,"

    So, it's only going to last 200 months (over 16 years). Perhaps Windows 95 is a better solution.
  • I'm guessing the issue is rosetta and non-native processes. I don't have any scientific evidence to back it up, but it makes sense that a square peg isn't compatible with a round hole. Sure you can make that hole big enough to put the square peg in but then your previous round peg just won't fit the same way.

    My employer went balls to the wall when they ordered 28 new M1 laptops last December. They were supposed to be the cure all godsend for everyone who thinks their current MBP was slow or didn't offer eno

    • Why would non-native processes write more to the file system?
    • Thats nonsense, once rosetta has translated the x86 binary to m1, the remainde rof the lifetime of the cpu there shouldnt be extra reads or writes. The tralsation is the only time extra writes should happen.
  • by Joe_Dragon ( 2206452 ) on Tuesday February 23, 2021 @11:24PM (#61094586)

    low ram = more swap and the high speed io hides slow down of that low ram.

    • Why would there be more ram consumed ? x86 is hardly a winner in terms of compact instruction encoding if anything its almost a guarantee tht the arm equivalent will actually be significantly smaller. If anything you shoul dhave more free mem after loading the same program on either on arm (m1) than on x86.
  • I can't imagine that's good for the battery...

  • 1. Deny, Deny, Deny.
    2. Hope bad publicity goes away.
    3. Acknowledge issue may affect some users.
    4. Hope bad publicity goes away.
    5. Fix for a subset of users.
    6. Hope bad publicity goes away.
    7. Acknowledge the issue affects everyone.
    8. Hope bad publicity goes away.
    9. Fix for everyone.
    10. Hope bad publicity goes away.
    11. File for bankruptcy.
    12. Hope bad publicity goes away.
    13. Move to the Cayman Islands.
    14. Hope bad publicity goes away.

  • by MemoryDragon ( 544441 ) on Wednesday February 24, 2021 @03:00AM (#61094852)

    A ram and swap issue.
    The base configuration of the m1 macs is sold with 8 gig which was way to few ram to avoic excessive swaps for even normal usecases.
    Never skimp on ram, rule #1!

    • by ledow ( 319597 )

      While I agree, I ran 4Gb + SSD machines for YEARS, literally almost a decade, without seeing significant hits to their TBW - they're still at 99% or 98% lifetime, and are in multi-user offices doing... well... all the office tasks you would expect.

      I suspect it's far more likely to be configured to hit the disk much more quickly and/or to cache things to SSD by preference, where even a little bit of tuning would drastically reduce the number writes.

      If not, then it's doing something VERY bad. This is either

    • A ram and swap issue.

      I admire the amount of confidence that you show. The problem is that what we saw reported is totally ridiculous numbers. Not explainable with a "ram and swap issue". I wouldn't dare giving any diagnosis without having one of these machines in my hands and examining them.

  • I don't know why people are shocked at this, it was clearly obvious from day one that they didn't put enough RAM in the machines and so now they are swapping like crazy. This is exactly why I won't buy a first generation M1, I'm not touching them until I can get at a very minimum at least 32GB of RAM. I have 128GB of RAM in my Mac Pro 2012, so even just 32GB would be a major step down.

    • by Mal-2 ( 675116 )

      Chromebooks tend to thrash with 2 GB, and are generally fine with 4 GB. There is just some tipping point where things go from "not bad " to "totally unusable" very quickly. It just can't be papered over by a fast SSD (generally, that is, there are some Chromebooks that shipped with surprisingly fast storage like the Acer C720 which easily saturates the SATA III bus). It looks like the whole cross-platform thing with the M1 and OS X may be just enough extra load to make 8 GB machines fall below the usable po

  • There are two possibilities: 1. The SSD drives are used much more than they should. 2. The code measuring the use of SSD drives is somehow wrong.

    Wait a few weeks. If the reports disappear then it was a false alarm. If not then it is real.
  • by Gabest ( 852807 ) on Wednesday February 24, 2021 @05:24AM (#61095042)

    Like the battery, keyboard, lcd. They will die in 5-10 years. I have replaced all of them in my notebooks from ebay two or three times before moving on to the next.

  • Comment removed (Score:5, Interesting)

    by account_deleted ( 4530225 ) on Wednesday February 24, 2021 @05:45AM (#61095076)
    Comment removed based on user account deletion

Philosophy: A route of many roads leading from nowhere to nothing. -- Ambrose Bierce

Working...