> I bet it would make a very formidable low power desktop

Perhaps if Apple really transitions to ARM on the Mac Line, that might be a reality in a couple years. I'm not sure they would want to replace x86 for the 'Pro' Desktop machines, but I don't think an ARM Mac Mini is out of the question.

ARM for Mac depends upon the Software ecosystem. Though much of enthusiast tools are ARM compatible, in the Linux/ARM scene; professional software are long way from it.

Unlike a dev board, Mac caters to professional audience. Apart from iOS apps, Several professional x86 software needs to be ported to ARM before ARM Mac becomes a reality.

If Apple decides to switch to ARM chips, they will have some form of virtualisation technology that'll allow unported apps to run.

Sure, performance won't be as good as native, but it'll be good enough to introduce ARM Macs.

If they don't have a way to emulate x86 chips, I don't see them introducing ARM Macs.

Yes, Mac ARM chips are likely to feature some x86 emulation.

On a side note, I wonder about the patent dynamics regarding x86 & x86-64 emulation.

When Microsoft & Qualcomm announced snapdragon powered Windows 10, intel gave a 'not-so-subtle' threat to Qualcomm regarding patents for x86 emulation[1]. MS, still released the model anyways.

I guess, Apple has enough leverage to negotiate patents for x86(intel),x86_64(AMD) emulation.

Both Intel and ARM will tell you that their patents are required.

(Questionable if that's true.)

> x86 emulation[1]

Did you mean to post a reference?

The main points of Apple changing to ARM is:

1. They don't need to pay licenses per core, so they can put 48, 96 cores in a single machine.

2. Is ARM as powerful as Intel, of course not, but who cares? ARM is more efficient so just add more cores to it.

3.Apple already owns all development stack. They already created hardware emulators for previous chip changes. They will do it again.

> so they can put 48, 96 cores in a single machine.

> so just add more cores to it.

Oh my...thermals. I don't know where to begin. Scaling cores (even if they're efficient) poses many challenges - thermal, core-to-core bandwidth, interposer design issues and warpage, and most importantly, yield as the die size gets larger. Or you'd have to break up the core dice into many die and create fabric such as what AMD Epyc server chips are doing, etc...etc. A 96 core die with decent power envelope would be in the order of 300-500 watts.

Is (2) true still? I thought I'd read that Apple's ARM chips were now somewhat competitive performance wise with Intel's chips?

There is no shortage of software to run on a sufficiently fast ARM processor. [0] The bigger problem is that (even without emulation) most ARM processors aren't powerful enough in comparison to x86 processors. Apples processors are the exception and they are only available with locked down operating systems.

[0] https://youtu.be/T-pPkpIDNnM?t=388

> Apart from iOS apps, Several professional x86 software needs to be ported to ARM before ARM Mac becomes a reality.

While this is true, the iPad Pro is the key. If Apple can continue to get pro apps ported to the pro they will have many already moved to ARM by the time a full machine comes out. See many Adobe apps already out or in progress.

Apple uses more die area for their chips. So it wouldn’t be cost competitive with what Qualcomm and Samsung ship.

This is it. Apple follows a mobile methodology for their chip designs (tall: very powerful, large size cores) while Qualcomm, Samsung, etc follow an embedded methodology (wide: many small efficient cores). On top of that, Apple has some great CPU engineers who've worked on the 68k and POWER architectures which they've leveraged against their ARM design. They're also the only consumers of their CPUs which allows them to design them against their use cases and optimize for them exclusively.

While I'm a big fan of ARM in general, the only other chips really competing against Apple in it's performance + efficiency are nVidia's Denver cores and the ThunderX2; for similar reasons. Not that Qualcomm or Samsung's chips are bad by any means; just different.

I wish Qualcomm took all that legal revenue and converted it into a fab. They are basically limiting their upside now. Patent licensing for wireless standards has to be "fair". If the LTE/5g patents were open, they could charge the max possible amount for a fully integrated arm + modem core optimized down to the material science.

Would be amazing if Apple made an ARM Xserve :D

They would probably be good for servers if Apple cared.

Apple has a LOT of servers. They just don’t sell them (well, they do). But they care a lot about it; they hired the Netty team to work on fundamental server software stacks (which we know because it’s open source). If ARM servers mean more power efficient data centers then I can see them doing it just for their sustainability objectives which is a major priority.

If an Apple-developed server SOC was put into wide use internally for sustainability reasons, wouldn’t Apple be morally obligated to open source such a design?

It seems like a possible conflict if it had tech otherwise proprietary to the A series.

where would this moral obligation or possible conflict come from?

Reading the comment again, I understand the possible conflict (they wouldn't open source proprietary designs), but still not the moral obligation.