I am not here to defend Apple. God knows, they have not given me anything I have not paid for (other than Apple TV+), I haven’t a single Apple share, nor is my dwarf cactus a distant relative of Steve Jobs. I am just a tech enthusiast like many others, but perhaps with a bit more balanced view on the tech world than customary these days — hence my statement that all this “SSDs are dying because of M1” is an overreaction, and tech news outlets publishing sensationalist articles with zero evidence is dumb, dangerous and goes against the very essence of what journalism is meant to be about.
Over the last three months I have paid very close attention to what some would call #SwapGate. It started with some YouTubers trying to justify the choice between 8GB and 16GB RAM, or 256GB and 512GB SSD. I would not see their intentions as malicious, but rather poorly educated, which begs the question — why?!? Why is YouTube split into two camps — one that claims there is no SSD swap issue, and another that claims there is. Same for Twitter. One guy posts a screenshot, and a day later tends if not hundreds of news outlets share that one screenshot, creating the impression that we’re dealing with a widespread issue. Needless to say, I am not going to share that same screenshot!
Naturally, owning an M1 since its launch, the concerns raised seemed legit, and I still believe they are. But having a legitimate concern does not equate to having an actual problem. Hypochondriacs have legitimate concerns all the time, but rarely do those turn out to be actual illnesses. I think we can also all agree that being concerned about the possibility of a nuclear war is very different than being in one — and let’s hope, we’ll never have to prove that. Same goes for the M1 and its alleged “problem” and associated concerns.
An analysis of the problem proposition
SSDs have a limited life-span. This is a fact. But this is also a fact for HDDs, and believe it or not, the average life-span of a spinning-disk drive is shorter than an SSD’s — by a lot! I am not going to regurgitate what others have already explained in intricate detail, just going to highlight that the average life-span of a spinning HDD is 5–7 years. That largely depends on how you use it though. Leaving a HDD running 24/7, will likely kill it in the space of 2–3 years.
By comparison, SSDs (solid state drives) have a longer average life-span, and it’s considered to be around 10 years. Having said that, leaving them always running is calculated quite differently. For all intents and practical purposes an SSD can be read an “infinite number of times”, but the number of times it can be written to is finite and it’s calculated in TBW (terabytes written) or — and this is very important — years, whichever limit is reached first. But it’s important to keep in mind that this is the manufacturer’s warranty period. We have all used technology long enough by now to expect most of our devices and appliances to last much longer than their warranty period. When you buy a car, its warranty states x number of miles or years, whichever limit is reached first. In an electric vehicle, the battery has its own warranty of x number of charging cycles or years.
Both HDDs and SSDs have something in common though, regardless of very different technologies being used to store data: usage pattern. Users writing massive amounts of data to the drive regularly, will wear out the drives quicker. This does not however mean that reaching the warranty limit renders the drive dead and since the introduction of SMART diagnostics for drives, preventing an unexpected catastrophic failure is much easier than it used to be before — though not an absolute fail-safe prevention method (always back up your stuff!).
Knowing all of the above one can see where the initial concerns came from. With a smaller RAM size and SSD size, one can make an educated enough guess and assume that the life-span of their drive will be shorter than it used to be before. This is what I think, is a fair assumption backed by historical evidence. However, note the operative word there in the subheading is: historical and in tech that can matter more than in many other industries.
SSDs over the last number of years have improved by a lot, and I mean exponentially. The TBW has gotten higher and higher, speeds improved to levels we never dreamed of, wear-leveling has improved, and cost has gone down by a lot!
Another historical fact is that Apple’s SSDs are famously reliable. I went through many Macs (around 10 of them) in the last 7 years. The drives were never an issue. From anecdotal evidence the most common issue seems to be the batteries expanding over time. Boom!💥
Apple’s SSDs don’t have an advertised TBW. Tech conspiracy theorists tend to latch on to this like a chihuahua on to the postal officer’s pants. To be the devil’s advocate for a second though, it can be seen as suspicious, but then looking at the company’s track record of not advertising overly technical details being the norm, becomes a lot less so. If the thing does the job, and does it well for many years, customers are happy, so TBW becomes irrelevant to 99.9% of customers.
Some users in the past have had bad experiences with SSDs. A lot more people have also had bad experiences with HDDs. That is partly why we now have SSDs! Just like not every user is the same, not every drive is the same, so it’s not surprising that we all have individual concerns and would like to address those. But unfortunately all we have is limited evidence in even more limited context — hence labelling it anecdotal. To illustrate, I’ll use my own M1 Pro with 8GB RAM, and 256GB SSD as an example, and it’s a great example because it’s the exact configuration “everyone” is so stressed out about.
When I bought this machine, I knew it was a gamble to some extent, but I genuinely wanted to see if the lowest spec M1 Pro machine can replace my current daily-driver — a 15" Pro, with 16GB RAM and 512 GB SSD or at the very least test out its reasonable limits. Not so much for myself, but to have first-hand proof of whether that is the case or not and know what to recommend later to friends and colleagues for various types of work like general software, web, mobile development or audio-video work. As a worst case scenario, it was going to replace my entry-level Air that I traded in for the M1 Pro, and I knew for fact before even powering the M1 Pro up, it will do that and a lot more. After all, the Air was just powerful enough for mid-complexity web development. Everything else beyond that, it tended to struggle with enough not be able to consider it great UX.
Now let’s do what all the tech journalists out there failed so very miserably to do, and analyse the output above in actual context. I purchased the machine on the 19th of November 2020. As I am writing this paragraph on the very same machine, it is the 19th of March, which makes this an exactly 4 month old machine.
- 294 power on hours = 42 days of power-on time at 7h / day.
- 188 power cycles because this machine (and frankly all my machines) is rarely ever shut down, I mostly just put it to sleep.
- 66.9 TB Data units read, but who the heck cares, reads are essentially infinite anyway.
- 28.3 TB Data units written, so 0.66 TB / day.
You can see the emphasis is on that last bit, and what so many people are suddenly freaking out about. Those 28.3 TB written represent 2% of the TBW of the drive according to SmartMonTools, released on the 30th of December 2020. Allow me to reiterate, so it doesn’t get missed in translation:
A piece of software at a version barely a month older than the Apple M1 machines, reports that I have so far, in exactly 4 months, used up 2% of my SSD’s manufacturer-warranted life.
I am being somewhat deliberately patronising here, because while I am terrible at maths, I am able to calculate that 2% usage in 4 months, means 6% in 12 months, which last time I checked, meant a whole year. At this rate, this drive will last 16 years, and still not be void of warranty — based on TBW numbers alone, that is.
But 660 GB written in a single day, every day?!?
Ehh. Yes, and here’s what I do to make that happen, which is very important to take into account because the apps you run can make a huge difference in how your drive is used — another “detail” tech journalists didn’t care to research.
I am going to pull the classic “trust me, I’m an engineer” line here, except on this occasion I am actually being serious, because I am a software engineer, and I do spend quite a lot of time on this machine, coding. WebStorm, Android Studio for Flutter development, VSCode, Xcode are tools I often use to work on various projects of my own.
In terms of apps that are always open, it’s Safari, Apple Music, iMessage and Signal, with LittleSnitch and 1Password always running in the background. On this machine I also work a lot in Apple Motion 5, did a whole course on it before that (so plenty Udemy streaming), do some audio editing with Reaper, iZotope and Audacity. Occasionally I’ll spin up MS Teams or Outlook and Affinity Photo. I do all my Zoom and Skype calls on this machine and I do watch YouTube and write Medium blogs on it as well.
It’s also worth noting that I keep the OS up to date, download all the new versions, and that includes updates to the apps. It’s also permanently connected to Time Machine to back up whenever it needs to.
There are three very important aspects to remember here:
- I use this machine for a wide array of tasks, and I do use it a fair amount.
- Not all the apps I use are M1 optimised. Wanna check? Head over to isapplesiliconready.com and it’ll tell you all you need to know. There may very well be a correlation between data writes and the use of Rosetta 2 for translating Intel apps to M1 architecture.
- I do not use Chrome. Not on this machine. I made a point of that. It is installed, and occasionally I fire it up just to debug something in a web app or website I am building, but no actual web browsing is done, all happens locally on the machine.
Take all that into account, and the daily 660 GB of written data start making a bit more sense. Having said that, some might say, that’s still too high. Well, let me tell you a story…
Back in 2007 when Windows Vista started to become popular, a lot of PC fans (believe it or not I used to hate Macs), including myself, were outraged when we realised that Vista, unlike XP, used all the RAM it could get its hands on. But then my friend — who was equally outraged — and I started thinking. If I had 2GB of RAM back then, and the operating system took no advantage of that, wouldn’t that have been a waste of resources? The more we thought of it, the more it made sense what Microsoft decided to do: use as much of the resources as you can whenever you have it. Within safe parameters of course, to allow all the other apps competing for memory as well.
Back to present day, when SSDs are lightning-fast, it starts to make more and more sense to use it as more than just storage, but also as a RAM extension. But again, within safe parameters, and this is practically the biggest take away of this article, and what should have been everybody else’s conclusion as well.
Technologists were right to pop their heads up looking at the new M1 machines and how RAM, SSD, GPU all together create a different hardware climate going forward, because that is indubitably the case. We are now dealing with a different kind of monster than with the Intel CPUs, but different doesn’t equate to bad. It can be proven, and has been a few times now, that the M1 machines that come with 8GB of RAM, will use more — if not double — the swap space on an SSD, and that will without a doubt cause more writes to the drive over the course of its life, but that amount will only mean the drive’s TBW life will shrink to 16 years versus 32 years, neither numbers representing anywhere near a practical, real-life use-case of a commercial entry-level laptop.
But let’s say you’re the absolute edge-case that needs their laptop to run and do stuff 24/7. Then having the configuration I have, would result in reaching the SSD’s warranted life in 5 years. In my book, under those circumstances, even that is a very respectable amount of time, and nothing to get our knickers in a bunch over. And remember, those 5 years are the warranted years, you could very well get a good few more out of that drive.
But it’s soldered to the motherboard…
Uh-hum. While I do empathise and to some extent support the Right to repair sentiment, I am not as outraged by the lack of removability of a memory chip by the user or some street corner computer-repair shop. I used to work in one right out of high-school, and let’s just say I’ve seen far too many things go wrong. But in my mind the argument around a user being able to replace the drive themselves doesn’t hold much water, not because it wouldn’t be great, because it would, but rather because the expectation is based on tradition rather than a genuine need.
Historically speaking we’re used to being able to replace drives all the time. I used to keep my PC open at all times, because I’d swap drives on a regular basis for a number of reasons. But that was me — the PC enthusiast and later computer technician. Replacing drives was also much more of a need back in the day, than it is today. Back then people would upgrade more often, or have a failed drive much more often, especially in laptops. The largest market for upgrades and drive replacements was teenagers and young adults, and occasionally a middle-aged PC enthusiast. Believe it or not, that is a fairly small market in the grand scheme of things, and now even smaller because laptops are so commonplace that barely anyone but enthusiasts and hard-core users and gamers need anything else. Times have changed, and user needs are different.
Additionally, there’s the aspect that few talk about — smartphones. You’re wondering what does that have to do with my Mac’s SSD? Well, a lot. I don’t see anyone banging on the table screaming for memory to be replaceable in a smartphone. Sure there’s folks who are advocates of external SD card storage, but that doesn’t mean the drive iOS or Android runs on is not soldered on to the motherboard. And that has been the case from before the smartphone era! No widely commercially available mobile phone had its memory replaceable ever. So if it’s fine in smartphones — the very devices we use 24/7 — why is it such a huge problem in laptops we use say 7h/day on average?!?
There is no verdict…
Yeah. The “it’s not an exact science” applies here very well. Just like it’s hard to bet on the weather, it’s also hard and ill-advised to bet on a piece of automated software that throws out some numbers that may or may not be accurate. Truth is, SMART data generated by software based on some drive tests is somewhat dependent on every manufacturer being honest with the numbers they claim as under warranty for their drives. Chances are, Apple has not yet provided that data, and it might never do, as we know, they’re not huge fans of advertising numbers that mean nothing or little to the end user. Therefore all the above data — my anecdotal evidence — and everyone else’s could be a big pile or worthless nonsense.
While this makes drawing conclusions a lot more difficult, it does illustrate the problem of too many variables to consider throwing out unverified and poorly researched claims on the interweb. Twitter is not a scientific journal, and even that is full of theories that can be refuted in next month’s issue. Irresponsible journalism relying on social media is the last thing the tech world needs in a society already saturated with fake news and “alternative facts”.
Attila Vago — Sr. Software Engineer building amazing ed-tech software. Cool nerd since forever, writer of codes and blogs. Web accessibility advocate, Lego fan, vinyl record collector. Loves craft beer!