Apple A14 - 5 nm, 11.8 billion transistors

[Eug]

A14 - First announced for iPad Air.

 

[Doug S]

Sounds about right. BTW, this was from August:

Foxconn starts major hiring drive for 'iPhone 12' assembly lines

The recruitment drive is an annual fixture of the Apple production schedule, with Foxconn typically hiring more staff in the months before the September launch of a new iPhone model. This year's call is later than 2019, when the recruitment drive commenced in July.

However, the chips would have to come out a bit before that of course.


Really? I had thought it was considerably less than 3 months.

EDIT:

I just looked it up. 6-8 weeks for simpler processes, but 3 months is about right for 7 nm.

They can run a so-called "hot lot" and get it faster (maybe fast enough for that 6-8 weeks) but you only do that on a new design to get test chips, find/fix bugs then run another hot lot etc. Apple probably had first silicon of A14 late last year, since it was reported it taped out in September.

 

[defferoo]

If someone buys a Macbook because it works well in Apple ecosystem or because it runs iPad apps (or something very similar) - he probably already own a Macbook.

I can't imagine anyone buy a Macbook over a Windows laptop or vice versa with security being a key factor. These systems are too different.

That was the point of my argument. We know for sure that some people will pass the next Macbook because their life on x86 will be much easier.

But it's really hard to point a scenario where someone would buy his first Macbook BECAUSE it moved to ARM.
So the outflow will dominate. That's it.

Again: that was not the argument.
You're talking about potential advantages of ARM over x86 (or Apple over Windows+Intel/AMD).

But most of the people you describe probably already have a Macbook.

Exactly.

the point others are making is not that the switch to ARM will make people jump on the Mac bandwagon, it’s that it could result in having a significantly better product, therefore more people buy Macs as a result.

for example, if a Mac runs all the applications you need twice as fast with longer battery life compared to other computers in the same price range, why would you opt for something else?

 

[piokos]

the point others are making is not that the switch to ARM will make people jump on the Mac bandwagon, it’s that it could result in having a significantly better product, therefore more people buy Macs as a result.

for example, if a Mac runs all the applications you need twice as fast with longer battery life compared to other computers in the same price range, why would you opt for something else?

Why would you not have a Macbook today if you can afford the price, like Apple's ecosystem and can live with software already is available on the platform?

Apart from the iPhone/iPad apps (which you could already run on your iPhone/iPad), software available on ARM Macs will be a subset of software available on x86 apps.

 

[name99]

If someone buys a Macbook because it works well in Apple ecosystem or because it runs iPad apps (or something very similar) - he probably already own a Macbook.

I can't imagine anyone buy a Macbook over a Windows laptop or vice versa with security being a key factor. These systems are too different.

That was the point of my argument. We know for sure that some people will pass the next Macbook because their life on x86 will be much easier.

But it's really hard to point a scenario where someone would buy his first Macbook BECAUSE it moved to ARM.
So the outflow will dominate. That's it.

Again: that was not the argument.
You're talking about potential advantages of ARM over x86 (or Apple over Windows+Intel/AMD).

But most of the people you describe probably already have a Macbook.

Exactly.

You agree that MANY more iPhones are in use than Macbooks?
So what happens over time (and we are talking ten years, not the next three months) as more and more of the people whose only computing device was a phone start to see the value proposition in also owning a desktop/laptop because of the larger screen, keyboard, multiple windows?

In the past they might have bought a cheap Windows PC or a Chromebook. But everything I suggested starts to make buying a Mac a better option because it the excess value (from being part of the iOS ecosystem) is so much larger than just being part of the Apple ecosystem.

Additionally COME ON. You don't think Apple is going to run EXACTLY the same playbook as the Phone/iPad/Watch once they actually control the SoC? ...
The 2021 and 2022 MacBooks will look more or less as we expect.
But 2023, once the market is saturated, changes will start. We'll see the new MacBooks at maybe $1000 and last year's model sold at maybe $800. Once year later we'll have three generations available, new at $1K, old at $800, older at $700.
Meanwhile we'll also see what's essentially the guts of an iPad (remember, that costs only $330!) glued to a keyboard and sold as the MacBook SE at maybe $400. The internet will sneer that it's garbage because it comes with only 8GB of RAM and only 128GB of flash. But it will be a MAC (multiple windows, trackpad, keyboard) and it will compete massively with the garbage laptops and chromebooks at that pricepoint.

Apple's playing a new game now. You're assuming the PC market will behave like the PC market of the last ten years. I foresee a PC market evolving like the Phone market of the past 10 years. The numbers will look slightly different for a long time because Windows PC's are sold to businesses as a different type of product, one that Apple isn't interested in creating. But look at the CONSUMER PC market over the next ten years.

 

[name99]

Why would you not have a Macbook today if you can afford the price, like Apple's ecosystem and can live with software already is available on the platform?

Apart from the iPhone/iPad apps (which you could already run on your iPhone/iPad), software available on ARM Macs will be a subset of software available on x86 apps.

Because the price will NOT stay the same...
While the software ecosystem will become wildly larger (both existing iOS apps easily ported, and new developers attracted by the larger market.)

You guys are perpetually living the past, assuming the world won't change, and assuming that anything Apple does was dreamed up a week before it happened, rather than being the fruit of an integrated five year plan all aspects of which were figured out in advance -- not just the HW and SW, but also the marketing strategy, the pricing strategy, when new price points will be hit, blah blah.

Look at the HISTORY of iPhone. Look at the HISTORY of iPad. Right now, two days ago dammit, you saw the newest version of this with this year's Apple Watch pricing and models. But you lot all assume, no, the structure of the PC market was frozen in 1995 and will never change!

 

[Eug]

Why would you not have a Macbook today if you can afford the price, like Apple's ecosystem and can live with software already is available on the platform?

Apart from the iPhone/iPad apps (which you could already run on your iPhone/iPad), software available on ARM Macs will be a subset of software available on x86 apps.

FWIW, I know a couple of people who are waiting for a Arm, just because... well... Arm. There is actually some logic to this. While Apple has said they will continue to support Intel for quite some time, they also said that for PowerPC, and those PowerPC machines became worthless pretty quick.

I've personally been telling people if they are basic users of Apple applications and that sort of thing, it may pay to wait for Arm. OTOH, if they are running their businesses using specific Intel Mac workflows, then buy now with Apple Care and then upgrade to Arm after the warranty runs out (and their software gets updated for Arm).

Actually, it's not just Arm. It's also because they're expecting form factor changes. For example, I have a 2010 iMac and a 2017 iMac, and from the front they look EXACTLY the same. Actually, it works out for me, since I'm using the 2010 as a matching external monitor for the 2017. And the 2020 iMacs also look exactly the same from the front. Some people are just bored of the form factor and want something new. I know that may seem odd to people who hang out in the CPU forum here, but it's actually how some mainstream users think.

BTW, in terms of software: The only thing that concerns me is MS Office. My legal copy is 2016 and I'm not paying a subscription. I'm sure Office will be available on Arm Macs pretty quickly, but I'm afraid they will force me to go a subscription model. Mind you, it could probably run well with Rosetta, and with my current job, I have less need for Office anyway.

BTW2, I already have a 2017 12" MacBook, but the three things that bug me about it are the mediocre keyboard, the weak trackpad feedback, and the single USB-C port. While I'll probably just stick with it since I have 16 GB RAM and the performance is sufficient, I'd love to see one with 2 USB-C ports, improved keyboard, and improved trackpad, on an Arm platform, fanless with longer battery life and maybe even 5G support. That would be my dream MacBook. (I don't like 13" laptops as much.)

 

[piokos]

FWIW, I know a couple of people who are waiting for a Arm, just because... well... Arm.

And the don't own a Macbook right now? Why?

BTW, in terms of software: The only thing that concerns me is MS Office. My legal copy is 2016 and I'm not paying a subscription. I'm sure Office will be available on Arm Macs pretty quickly, but I'm afraid they will force me to go a subscription model. Mind you, it could probably run well with Rosetta, and with my current job, I have less need for Office anyway.

Again: I've specifically mentioned developers and scientists. This is a group that buys a lot of MacBooks Pro and which will be limited by factors other than just how much of modern software is available on ARM (or runs via Rosetta).
Because they use a lot of legacy programs written for x86. And because they develop for x86, so it's easier on x86.

I'm also fairly worried about modern software.
Some companies see Mac owners as core client group, so I'm absolutely sure stuff from Adobe or Microsoft will be available more or less instantly.
I'm not so sure about engineering/science software - say, Matlab. Matlab is so deeply optimized for x86, it only works on Macs because of hardware similarities (and there are still some issues not existing on Windows or Linux - meaning it's more a port than an actual solid implementation).

Because the price will NOT stay the same...

You know... sometimes I think you don't understand my arguments at all. And sometimes that you're Tim Cook in disguise.

If your only stance in this topic is to praise Apple, because you think I'm against them in whatever way, you're just wasting time.

 

[SpaceBeer]

Everything is moving to cloud. Whatever is missing, there will be option to access it via web browser - CAx (Onshape, SimScale...), science (MathWorks Cloud, Google Cloud...), gaming (Stadia, xCloud...), office (MS Office Online, Collabora...). The only problem users might face is, all of these services are optimized for Chrome But some might have simple desktop applications that can probably easily be ported to new architecture (e.g. Fusion 360).

Like it or not, ARM devices will take significant PC marketshare in following years. Not because they are better (they are usually not), but because Apple and Google wants that, and have resources (money) to make that happen

 

[teejee]

Everything is moving to cloud. Whatever is missing, there will be option to access it via web browser - CAx (Onshape, SimScale...), science (MathWorks Cloud, Google Cloud...), gaming (Stadia, xCloud...), office (MS Office Online, Collabora...). The only problem users might face is, all of these services are optimized for Chrome But some might have simple desktop applications that can probably easily be ported to new architecture (e.g. Fusion 360).

Like it or not, ARM devices will take significant PC marketshare in following years. Not because they are better (they are usually not), but because Apple and Google wants that, and have resources (money) to make that happen

Chrome will probably be available for ARM Macs

Chromium Docs - Chromium for Arm Macs

 

[beginner99]

for example, if a Mac runs all the applications you need twice as fast with longer battery life compared to other computers in the same price range, why would you opt for something else?

Because I hate closed platforms that limit your software options.

 

[name99]

And the don't own a Macbook right now? Why?

Again: I've specifically mentioned developers and scientists. This is a group that buys a lot of MacBooks Pro and which will be limited by factors other than just how much of modern software is available on ARM (or runs via Rosetta).
Because they use a lot of legacy programs written for x86. And because they develop for x86, so it's easier on x86.

I'm also fairly worried about modern software.
Some companies see Mac owners as core client group, so I'm absolutely sure stuff from Adobe or Microsoft will be available more or less instantly.
I'm not so sure about engineering/science software - say, Matlab. Matlab is so deeply optimized for x86, it only works on Macs because of hardware similarities (and there are still some issues not existing on Windows or Linux - meaning it's more a port than an actual solid implementation).

You know... sometimes I think you don't understand my arguments at all. And sometimes that you're Tim Cook in disguise.

If your only stance in this topic is to praise Apple, because you think I'm against them in whatever way, you're just wasting time.

Like I keep saying, you are thinking of this in terms of 1995; your view of "who owns a PC" consists of the US and European middle class. The whole point of my argument is that over the next ten years Apple will provide something of compelling value to people who have so far not seen that compelling value in traditional PCs (either the lousy value and hassle of Windows at $350, or the MacBook at $1000).

If you're going to resort to calling everyone who disagrees with your opinion a fanboi or a Tim Cook in disguise, then go play with the other teenagers. I don't have time to waste with people who can't get their heads around the idea that someone may disagree with them for what are actually serious and rational reasons.

 

[Doug S]

Actually, it's not just Arm. It's also because they're expecting form factor changes. For example, I have a 2010 iMac and a 2017 iMac, and from the front they look EXACTLY the same. Actually, it works out for me, since I'm using the 2010 as a matching external monitor for the 2017. And the 2020 iMacs also look exactly the same from the front. Some people are just bored of the form factor and want something new. I know that may seem odd to people who hang out in the CPU forum here, but it's actually how some mainstream users think.

How are they going to change the form factor of the iMac, when it by definition looks exactly like a monitor? That's the one model you can say for sure won't change (unless you consider having smaller bezels to be a "form factor change")

The other desktop models might change though. The Mac Mini could become the size of a deck of cards if they want, and who what happens with the Mac Pro - maybe the next one is a tall pyramid like the Transamerica tower lol

 

[piokos]

Everything is moving to cloud. Whatever is missing, there will be option to access it via web browser - CAx (Onshape, SimScale...), science (MathWorks Cloud, Google Cloud...), gaming (Stadia, xCloud...), office (MS Office Online, Collabora...). The only problem users might face is, all of these services are optimized for Chrome But some might have simple desktop applications that can probably easily be ported to new architecture (e.g. Fusion 360).

Like it or not, ARM devices will take significant PC marketshare in following years. Not because they are better (they are usually not), but because Apple and Google wants that, and have resources (money) to make that happen

Absolutely. My whole work is happening in the cloud. I could work on a Chromebook. In fact, if I was given the choice, I probably would ask for an ARM laptop next time (but it would have to be Windows or MacOS).
And I assure you cloud is better.

Personally I'm still mostly local, but I'm migrating as much as I can afford.

However, once you assume ARM is OK because you're cloud-based anyway, the argument about superior performance loses most of its charm.
That's what I'm trying to point out: the arguments that people use to support the idea are just a tad incoherent.

I have nothing against ARM (I already use it in RPis and cloud) and I definitely have nothing against Apple. But I'm not convinced by ARM laptops replacing x86 so early. ARM simply doesn't support existing workflows or forces a much more complex setup. And, while I'm sure this will change over time (and 10 years from now ARM and x86 may be interchangeable), migrating today seems like a very extreme idea.

 

[Eug]

How are they going to change the form factor of the iMac, when it by definition looks exactly like a monitor? That's the one model you can say for sure won't change (unless you consider having smaller bezels to be a "form factor change")

The other desktop models might change though. The Mac Mini could become the size of a deck of cards if they want, and who what happens with the Mac Pro - maybe the next one is a tall pyramid like the Transamerica tower lol

The iMac has a ginormous chin and huge bezels. People have been commenting about those for years. They have been the same size for literally over a decade. Yes bezel and chin size reduction is a form factor change.

Actually, the chin affects its ergonomics too since it raises the height of the machine more than is necessary. Personally I’d like to see height adjustability too but I’m guessing they won’t do that.

And the don't own a Macbook right now? Why?

Own a Mac but over 5 years old or else a different model. ie. Overdue for an upgrade but waiting for Arm or else someone with a desktop looking to add a laptop.

Yes I know where you’re coming from but personally I think restricting to switchers for such an argument is being too narrow.

 

[Eug]

I can't find these A14 Geekbench 5 scores on Geekbench's website, but nonetheless they've been published at CPU Monkey. Maybe it's because the Chinese YouTubers already have the iPad Air 4 available for review.

Apple A14 Bionic Benchmark, Test and specs

Apple A14 Bionic benchmark results and review of this cpu with specs including the number of cores, threads, memory bandwidth, pcie lanes and power consumption. Benchmarks in Cinebench R23 and Geekbench 5

www.cpu-monkey.com

1658 single-core
4068 multi-core
2.45X multiplier

Contrast that to A13:

iPhone 11 Pro - Geekbench

Benchmark results for an iPhone 11 Pro with an Apple A13 Bionic processor.

browser.geekbench.com

1337 single-core
3507 multi-core
2.62X multiplier

If these A14 numbers are legit, that means that A14 is 24% faster single-core and 16% faster multi-core as compared to A13.

As for A12:

iPad mini (5th generation) - Geekbench

Benchmark results for an iPad mini (5th generation) with an Apple A12 Bionic processor.

browser.geekbench.com

1117 single-core
2721 multi-core
2.44X multiplier

That means A14 is 48% and 49% faster A12 for single-core and multi-core respectively.

 

[IvanKaramazov]

I can't find these A14 Geekbench 5 scores on Geekbench's website, but nonetheless they've been published at CPU Monkey. Maybe it's because the Chinese YouTubers already have the iPad Air 4 available for review.

Apple A14 Bionic Benchmark, Test and specs

Apple A14 Bionic benchmark results and review of this cpu with specs including the number of cores, threads, memory bandwidth, pcie lanes and power consumption. Benchmarks in Cinebench R23 and Geekbench 5

www.cpu-monkey.com

1658 single-core
4068 multi-core
2.45X multiplier

Contrast that to A13:

iPhone 11 Pro - Geekbench

Benchmark results for an iPhone 11 Pro with an Apple A13 Bionic processor.

browser.geekbench.com

I'd love to believe those numbers, but there are some oddities. For example, it lists 2 Icestorm cores and 4 Firestorm cores. Those are the code names from the Bloomberg report regarding the 8x4 Mac chip, but with the names applied to the wrong cores. Could be a mistake on cpu monkey, but it smells a bit funny. In fact there's a lot of specific info listed in there that seems unlikely for them to have already teased out, but who knows?

 

[Eug]

I'd love to believe those numbers, but there are some oddities. For example, it lists 2 Icestorm cores and 4 Firestorm cores. Those are the code names from the Bloomberg report regarding the 8x4 Mac chip, but with the names applied to the wrong cores. Could be a mistake on cpu monkey, but it smells a bit funny. In fact there's a lot of specific info listed in there that seems unlikely for them to have already teased out, but who knows?

I think they may have filler info there. For example they list 16 GB max memory. Did they get that from the A12Z dev box? It’d be better if they just left those things blank.

GSM Arena does the same thing. They fill in the info from known facts and also from published rumours and then correct it later.

Still, they had to have gotten the GB5 numbers from somewhere, unless they just completely made them up, but AFAIK they are not known for doing that.

 

[IvanKaramazov]

Yeah, there's definitely some filler and that could explain the naming mistakes. I wondered about that 16GB as well, and the 8GB listed as "max memory" for the GPU, particularly strange in a shared memory architecture.

Worth noting that the single core number they list there, 1658, is exactly the same as the single core score from the dubious March "leak". Not sure if that lends credence or casts doubt, but there you go.

EDIT - In fact, the single core score on cpu monkey is the same as the March leak to the digit, while the multi core score on cpu monkey is almost exactly 40% higher than the A12 multi-core score listed for the previous iPad Air in Geekbench's charts.

 

[name99]

I can't find these A14 Geekbench 5 scores on Geekbench's website, but nonetheless they've been published at CPU Monkey. Maybe it's because the Chinese YouTubers already have the iPad Air 4 available for review.

Apple A14 Bionic Benchmark, Test and specs

Apple A14 Bionic benchmark results and review of this cpu with specs including the number of cores, threads, memory bandwidth, pcie lanes and power consumption. Benchmarks in Cinebench R23 and Geekbench 5

www.cpu-monkey.com

1658 single-core
4068 multi-core
2.45X multiplier

Contrast that to A13:

iPhone 11 Pro - Geekbench

Benchmark results for an iPhone 11 Pro with an Apple A13 Bionic processor.

browser.geekbench.com

1337 single-core
3507 multi-core
2.62X multiplier

If these A14 numbers are legit, that means that A14 is 24% faster single-core and 16% faster multi-core as compared to A13.

As for A12:

iPad mini (5th generation) - Geekbench

Benchmark results for an iPad mini (5th generation) with an Apple A12 Bionic processor.

browser.geekbench.com

1117 single-core
2721 multi-core
2.44X multiplier

That means A14 is 48% and 49% faster A12 for single-core and multi-core respectively.

Every year we get these supposed Geekbench numbers (multiple different versions) and every year most of them turn out to be fakes.

The numbers you are quoting
(a) turned up around March, which seems awfully early...
(b) you will notice they ONLY give the summary numbers. My take is that if a leaker isn't willing to give the *entire* set of numbers (so that one can examine patterns, and thereby ascertain plausibility) it's a waste of time even considering the numbers -- by far the most plausible hypothesis is that they were made up by someone seeking attention.

 

[scannall]

Every year we get these supposed Geekbench numbers (multiple different versions) and every year most of them turn out to be fakes.

The numbers you are quoting
(a) turned up around March, which seems awfully early...
(b) you will notice they ONLY give the summary numbers. My take is that if a leaker isn't willing to give the *entire* set of numbers (so that one can examine patterns, and thereby ascertain plausibility) it's a waste of time even considering the numbers -- by far the most plausible hypothesis is that they were made up by someone seeking attention.

Personally I look at 'leaked' numbers as amusement, not to be taken seriously. I am patient, and willing to wait for reputable sites actually testing real in the wild products.

 

[Eug]

Every year we get these supposed Geekbench numbers (multiple different versions) and every year most of them turn out to be fakes.

The numbers you are quoting
(a) turned up around March, which seems awfully early...
(b) you will notice they ONLY give the summary numbers. My take is that if a leaker isn't willing to give the *entire* set of numbers (so that one can examine patterns, and thereby ascertain plausibility) it's a waste of time even considering the numbers -- by far the most plausible hypothesis is that they were made up by someone seeking attention.

Oh these are the same March ones? Hmm...

The number I have issue with is the 1658 single core. I was guessing 15xx so this seems a tad too high. The 4068 seems reasonable though.

Personally I look at 'leaked' numbers as amusement, not to be taken seriously. I am patient, and willing to wait for reputable sites actually testing real in the wild products.

Well, the YouTubers have published (p)reviews already (albeit in Chinese). In this video you can see Touch ID integrated into the power button being tested:

But if the numbers are the same March ones then that negates that argument. Also, now that I think about it, I believe these were at an event Apple organized, so they wouldn't have been able to install Geekbench on them.

 

[Doug S]

If those A14 numbers are correct, then the 16% CPU performance gain reverse engineered from the 40% A12->A14 gain was for multithread.

If the clock rate is 3.1 GHz that gives them a 16% increase all by itself, so the remaining 8% would be IPC improvement.

If true, and we know the A14s going into Macs will be even faster, there's no way Intel's (or AMD's) fastest will be able to keep up.

 

[Eug]

Every year we get these supposed Geekbench numbers (multiple different versions) and every year most of them turn out to be fakes.

The numbers you are quoting
(a) turned up around March, which seems awfully early...
(b) you will notice they ONLY give the summary numbers. My take is that if a leaker isn't willing to give the *entire* set of numbers (so that one can examine patterns, and thereby ascertain plausibility) it's a waste of time even considering the numbers -- by far the most plausible hypothesis is that they were made up by someone seeking attention.

So I looked this up. These numbers aren't quite the same as the March leak.

The single-core is the same at 1658, but the March leak has 4612 as the multi-core score, whereas it's 4068 for the CPU Monkey multi-core score.

That said, as I mentioned before, the number I had a harder time believing was actually the single-core score, as it seems a tad too high.

If those A14 numbers are correct, then the 16% CPU performance gain reverse engineered from the 40% A12->A14 gain was for multithread.

If the clock rate is 3.1 GHz that gives them a 16% increase all by itself, so the remaining 8% would be IPC improvement.

If true, and we know the A14s going into Macs will be even faster, there's no way Intel's (or AMD's) fastest will be able to keep up.

Even hypothetical CPUs based off last year's A13 core could make for a really fast Mac chip, considering that it's already scoring 1340 single-core.

 

[name99]

So I looked this up. These numbers aren't quite the same as the March leak.

The single-core is the same at 1658, but the March leak has 4612 as the multi-core score, whereas it's 4068 for the CPU Monkey multi-core score.

That said, as I mentioned before, the number I had a harder time believing was actually the single-core score, as it seems a tad too high.


Even hypothetical CPUs based off last year's A13 core could make for a really fast Mac chip, considering that it's already scoring 1340 single-core.

Either way no suite of numbers means I have zero way to validate their reliability...
If you honestly ran GB5, then you can take a goddamn screen shot of the full results!
Any monkey can look up last year's summary numbers, pick a multiplier like 1.23, and, voila, faked GB5 numbers...

It's hard to tell with the multi-core numbers in particular. I still don't actually understand what Apple's target is with those cores. Is the goal mainly to save power (in which case they seem substantially more performant than they need to be, and getting better every year). Or is the goal to provide lots of small cores to train developers to write more aggressively parallel code? (Do many apps use so many of the small cores?)
Given that I can't tell what Apple's goals are, I feel little confidence in predicting year to year whether they'll do very little to speed them up, or aggressively speed them up.

 

[piokos]

Own a Mac but over 5 years old or else a different model. ie. Overdue for an upgrade but waiting for Arm or else someone with a desktop looking to add a laptop.

So they would not be moving to Apple because ARM, but merely upgrading from an older system - so not the case we're looking for.
And there's nothing wrong or rare about using a computer for 5 years. It's pretty normal.

Once again... people either haven't read my original post or don't understand it. Some, like @name99 went into full "you don't praise a Macbook, so you're an Apple hater" mode.

I precisely said that Apple market share will drop IMO, because moving Macs to ARM will repel more people than it will attract into the ecosystem (if any).
This was a pretty simple, behavioral observation (with some arguments) that I could make about a type of potato. But since it was about computers, some people went berserk, because it's just not possible for Apple to lose clients.
And I haven't even said this is would be a business mistake. They may still earn more in the end.