Apple's M1 Ultra Chip: Everything You Need to Know

[Sf844]

Uhhh. So now you want to run a web server on a Mac? If not, why bring that up?



I'm not sure you know what a Venn diagram is.



We get it: your pet use case runs best on x86. Enjoy!

(By the way: so does mine. A lot of my work involves running VS in Windows, and a lot of that is still x64-only. But I don't go around telling people over and over that their use cases are invalid.)

So out of hundreds of millions of professional customers all over the world, ranging from agriculture all the way to banking, my point of view is niche because, you imply, it’s not representative of the overall market.

It appears you think most of those professionals are better served by a computer that cannot run x86 software but can run Logic Pro and Final Cut plus web based services, which can run on x86 equally.

You’re so delusional one feels speechless. I’m done here…

 

[chucker23n1]

So out of hundreds of millions of professional customers all over the world, ranging from agriculture all the way to banking, my point of view is niche because, you imply, it’s not representative of the overall market.

Again, I assure you most of those professionals can work on an ARM Mac just fine. The use cases that absolutely require x86 do exist, but are a shrinking minority.

 

[Sf844]

Again, I assure you most of those professionals can work on an ARM Mac just fine. The use cases that absolutely require x86 do exist, but are a shrinking minority.

Edit: forget it

 

[powerslave65]

I’m done here…

Hot damn!

 

[name99]

Thanks for that pointer to eclectic light. When I find time, I’ll probably skim through that in more detail, but found this interesting:

The hunt for the M1’s neural engine

How can you tell when software uses the Neural Engine in an M1 series Mac? How much power does it use, and what is Espresso? Mysteries unravelled.

eclecticlight.co

I think the conclusion is that the neural engines are under utilized, which probably means they will be doubly under utilized in an Ultra. Which is a shame, because there’s a lot of power there.

I agree with your assessment that this is still early days for such tools, and the point for now is to make the hardware available so people begin using it. The benefit of having the ANE cores in the M1 is likely less for the M1 users of today and more so Using them is mainstream by the M3, or whatever.


Based on what I read in the link above, it sounds like most third party AI work is still hitting the CPU/GPU Rather than the neural engines. If that’s true, then they are mostly being used for features in the OS and probably won’t benefit from the additional cores. The OS features are designed to run a common set of functionality fast enough to be transparent. Doubling the cores won’t help unless they’re going to make certain capabilities only available on the Ultra Studio, which they won’t.

One exception might be that someone commented that FinalCut does ML on input footage— that’s something that might benefit by running faster.

Same for the media engines— so far it doesn’t seem like tools like Handbrake are fully utilizing the media engines, but I’ll be interested to see if they eventually do.

Yes, once one thinks about it, one can imagine uses for an NPU everywhere. For example people who use Channels DVR (or perhaps Plex? I don't know that as well) are familiar with comskip, which analyzes recorded video to find the ads. This is the sort of task that seems like it could be improved by adding some ML to the current mix of analyses (which include things like looking for fade to black, channel logo, sudden change in volume, etc).
But think about how many years it will take for someone in the open source community to start playing with such an idea, then getting it to work well, then optimizing it for the Apple ML API's rather than running as generic C code on a generic CPU...

 

[name99]

Every single leak from Gurman on the post M1 release rollouts has come true, along with the Jade-C core photos that clearly show the M1, M1 Pro (Jade-C Chop) M1 Max (Jade-C), M1 Ultra (Jade-2c) and one unreleased monster core labeled Jade-4C that's twice the size of the Ultra.

Since Gurman gives no technical details, "coming true" is hard to analyze.
For example I could certainly believe that there exist two-SOCKET boards within Apple, which have an M1-Ultra in each socket. Those are perfectly fine as development machines, for the OS and driver teams to get things working for future designs that have CPU an GPU split over, say, four separate chips.

Does that count as "coming true"? If it's never shipped, only Apple-internal? If it is eventually shipped, but as two sockets (hence with a pronounced performance delta when you have to move from one socket to the other) rather than as a single "unified" package?

 

[name99]

Mate, be real. All software is windows first but two niche media editing programs made by Apple. If you haven’t noticed this you’re either a DJ/Youtuber or related, Apple coder or bought a 2500 plus dollars machine to surf the web and send mails with a fisher price native app.

There is nothing pro about a computer that cannot be used for pro software or pro peripherals. And by pro peripherals I mean HDMI, fe, not whatever gibberish niche port nobody has nor needs at work.

I’m sure you could create tremendously efficient professional vehicles and equipment if you could modify the whole world ad hoc to serve them. That’s not the way things work IRL. The theoretical potential doesn’t matter, what matters are the tangible results, and x86 is the only name in town there.

This is akin to buying a 1400 dollars ipad pro (it’s incredibly fast!) with keyboard and mouse and spend half a day trying to copy and modify files from an USB, a job you could perform in 3 minutes with a 7 years old random laptop. Solutions looking for problems that don’t exist. One can charge his laptop at night, at the very least, but can’t get pro software developers to release a version for arm.

ALL SOFTWARE? Someone appears to be unaware of, at the very least
- console games
- mobile
- web apps
- data warehouses

I think what you mean to say is "all x86-Windows-first software is x86 Windows first" which, while true, isn't quite as exciting...

 

[chucker23n1]

Since Gurman gives no technical details, "coming true" is hard to analyze.
For example I could certainly believe that there exist two-SOCKET boards within Apple, which have an M1-Ultra in each socket. Those are perfectly fine as development machines, for the OS and driver teams to get things working for future designs that have CPU an GPU split over, say, four separate chips.

Does that count as "coming true"? If it's never shipped, only Apple-internal? If it is eventually shipped, but as two sockets (hence with a pronounced performance delta when you have to move from one socket to the other) rather than as a single "unified" package?

Either way, it'll be interesting to see what they do about the M1 Ultra.

If they do ship a two-package approach, that does raise quite a few questions. How is the RAM shared? Perhaps there's no unified RAM at all, but rather RAM on slots? But then the latency and bandwidth would be a lot worse. Maybe they do a hybrid approach where each of the two M1s Ultra has some RAM, but you can choose per-process to instead use additional RAM on slots. This would also answer how you'd go from 256 GiB to 1.5 TiB, assuming the next Mac Pro still offers as much.

What about the GPU? Does it appear as two GPUs in Metal? That'll lead to a lot of apps not supporting this, and you as the user not benefitting from the performance.

Etc.

Seems easier for some of this to be solved by a hypothetical M2 Max, which unlike the M1 Max doesn't have one but two interconnects, so you can go for a 2x2 grid.

ALL SOFTWARE? Someone appears to be unaware of, at the very least
- console games
- mobile
- web apps
- data warehouses

I think what you mean to say is "all x86-Windows-first software is x86 Windows first" which, while true, isn't quite as exciting...

I agree, but technically, two out of three major consoles right now are x86. So most console games from the past decade are presumably x86, too.

 

[sudo-sandwich]

I put pepper on my cereal but don’t expect kelloggs to do it in the box.

Just because you use HDMI doesn’t mean it’s popular among Mac users anymore. There are plenty of Thunderbolt to HDMI converters, no need to add it to Macs and jack the price.

Considering how many products use HDMI, I don't think I'm the odd one out. TB to HDMI adaptors never work reliably, and it's bothersome to carry them around.

Don't worry, it doesn't really affect the price. If it even costs significantly more to add the port, that's just coming out of Apple's margins. They're not selling MBPs at-cost.

 

[sudo-sandwich]

But it is. After power and USB, it's probably the most commonly used port.

Yeah, and when not at a desk, it might even be more commonly used than USB. Heavily dependent on how many people still use flash drives.

 

[sudo-sandwich]

Either way, it'll be interesting to see what they do about the M1 Ultra.

If they do ship a two-package approach, that does raise quite a few questions. How is the RAM shared? Perhaps there's no unified RAM at all, but rather RAM on slots? But then the latency and bandwidth would be a lot worse. Maybe they do a hybrid approach where each of the two M1s Ultra has some RAM, but you can choose per-process to instead use additional RAM on slots. This would also answer how you'd go from 256 GiB to 1.5 TiB, assuming the next Mac Pro still offers as much.

Isn't this just the typical NUMA problem? On multi-socket machines even with the usual off-chip RAM, each node gets its own set of RAM. They can also access each other's RAM, but accessing their own is fastest.

In a lot of cases it'll work out, cause often times a process scaling well to 20 cores isn't sharing much memory anyway. Might as well use a cluster, but single machine is just easier to deal with. And processes using fewer threads are probably not scheduled across nodes.

 

[chucker23n1]

Isn't this just the typical NUMA problem? On multi-socket machines even with the usual off-chip RAM, each node gets its own set of RAM. They can also access each other's RAM, but accessing their own is fastest.

In a lot of cases it'll work out, cause often times a process scaling well to 20 cores isn't sharing much memory anyway. Might as well use a cluster, but single machine is just easier to deal with. And processes using fewer threads are probably not scheduled across nodes.

Yes, it's basically a form of NUMA, but I think it's a bit of a tricky sell because the M1 Pro, Max and Ultra presentations were so focused on high memory throughput.

 

[dogstar]

I mean… there was really not much of a meaningful difference between ActiveX and NPAPI, just that one was only used by Microsoft and the other, originally by Netscape, was also implemented by other browsers.

Way back in the day it was more intended to compete with Java beans as little apps running on the web page. You could build a little calculator or a game on a web page in activex lol. We were so naive back then..

 

[SlaveToSwift]

Considering how many products use HDMI, I don't think I'm the odd one out. TB to HDMI adaptors never work reliably, and it's bothersome to carry them around.

Don't worry, it doesn't really affect the price. If it even costs significantly more to add the port, that's just coming out of Apple's margins. They're not selling MBPs at-cost.

You think Apple throws anything in a build at cost? Every Mac is priced to achieve a specific gross margin percentage, typically roughly 50% so their is enough profit to pay for development, testing, marketing, sales, administration, rent, research, etc, etc.

So $40 in parts that requires $10 in assembly adds $100 to the base price of every single Mac. Any cost over $25 potentially bumps it $100 since Apple doesn’t price Macs anywhere other than at $100 increments.

 

[sudo-sandwich]

You think Apple throws anything in a build at cost? Every Mac is priced to achieve a specific gross margin percentage, typically roughly 50% so their is enough profit to pay for development, testing, marketing, sales, administration, rent, research, etc, etc.

So $40 in parts that requires $10 in assembly adds $100 to the base price of every single Mac. Any cost over $25 potentially bumps it $100 since Apple doesn’t price Macs anywhere other than at $100 increments.

This isn't the case, for the same reason that corporate taxes aren't "passed onto the consumer" 1:1. The cost is split in a complicated way. Despite Apple's current margin being maybe 100%, a $1 cost increase doesn't mean a $2 price increase, not even a $1 price increase. They're already charging as much as they can "get away with."

Regardless, the port probably only adds like $5 overall cost. You can even buy an adaptor, at retail price, for around that.

 

[sudo-sandwich]

This is akin to buying a 1400 dollars ipad pro (it’s incredibly fast!) with keyboard and mouse and spend half a day trying to copy and modify files from an USB, a job you could perform in 3 minutes with a 7 years old random laptop. Solutions looking for problems that don’t exist. One can charge his laptop at night, at the very least, but can’t get pro software developers to release a version for arm.

Given how much software is web-based and how much pro software runs on Mac already, with or without Rosetta, this is a different thing. The category of Windows-only software is constantly being eroded. In software engineering, most stuff is Windows-last. In creative pro fields, it's a mix.

iPads aren't even solutions looking for problems. They're problems, of the first-world variety. You're totally right that a random 7yo laptop (like the one I'm typing on) is usually easier because it's a better human interface regardless of whatever tech.

 

[EdT]

Given how much software is web-based and how much pro software runs on Mac already, with or without Rosetta, this is a different thing. The category of Windows-only software is constantly being eroded. In software engineering, most stuff isn't even Windows-first. In creative pro fields, it's a mix.

iPads aren't even solutions looking for problems. They're problems, of the first-world variety. You're totally right that a random 7yo laptop (like the one I'm typing on) is usually easier because it's a better human interface regardless of whatever tech.

iPads for me are portable entertainment devices (mostly shows and music) that in a pinch I can do Basic stuff acceptably fast and more involved things I can do but would have been much faster with a laptop.

I have one because when traveling I have a company Windows laptop that I’m not supposed to install “personal“ files onto. My phone does a good enough job with personal emails, my company computer handles work related software and my iPad has movies/downloaded streaming and music stored in it. Many areas that I am sent are still in the very early 2000’s as far as internet access is concerned, so I don’t depend on having decent access.

 

[DeepIn2U]

I completely understand. The bulk of the crowd can be so anti-Windows that they don't... or don't care... but those of us who have had a genuine NEED for Windows have enjoyed the FANTASTIC benefit of having a full Windows PC and a full Mac in a single box- desktop or laptop- since Apple went Intel. That was a HUGE benefit so readily assumed away by those who don't share that need and can do everything they want to do within macOS only.

For some of us, Windows is essential (for work or clients) while Mac was nice to have but not essential. And for some of that type, when their Intel Mac conks, they may face an either-or decision instead of both and have to go full Windows to cover need, vs Mac to cover want.

One positive about Windows is that it generally doesn't deprecate hardware as fast as Apple does. So if you have an Intel Mac on which you can install Windows 10, you should be covered for just about all things Windows through 2028. That option should let many like us put off having to buy a new PC for at least a while.

On the other hand, if you must have Windows 11, it's pretty much a 2-computer purchase scenario or only 1 (and that 1 will be Windows).

In my own case, that working iMac is conking out, taking my best Mac, my best PC and my best monitor with it all at the same time. So even before the last event, I was already mentally committed to not buying another "all in one" anyway and was already thinking Mac Mini with M1 MAX and smallish-sized Windows NUC/Ryzen (Mac mini-sized PC) and I had been wanting an ultra-wide monitor for a long time anyway. So I let the near-death of my trusty iMac spur on a few purchases to set up the new world of separates (computers) and a single ultra-wide for BOTH of them.

PC makers are building some very powerful PCs in surprisingly small boxes. I suspect some people who need PC but want Mac might flip the laptop concept into carrying 2 little PC boxes in a bag and use hotel TVs or client monitors when traveling... or perhaps buy a portable laptop-sized monitor to cover the screen base too. Yes, that would be a heavier bag than a single Intel MB able to run both platforms in a light-weight case. But, that would basically make full desktop power be "with us"... kind of like a couple of bigger, heavier dongles.

Why not consider using Windows via Azure hosting?! It’s not expensive and file transfer is pretty fast depending on your internet speeds.

Also hasn’t Windows for Arm (still in beta) not work with your applications even if you specifically set them to run in compatibility mode within the windows environment)?

There are options and I’ll bet windows in azure will work for you. This isn’t running virtual box on Apple silicon btw.

 

[polyphenol]

Why not consider using Windows via Azure hosting?

It is fine to consider it.

But there are problems such as dongle software protection. And some local hardware devices. Such as wireless connectivity to machines.

And I went to look for prices and found, as so often, thousands of options but not a simple price for an instance.

 

[HobeSoundDarryl]

Why not consider using Windows via Azure hosting?! It’s not expensive and file transfer is pretty fast depending on your internet speeds.

Also hasn’t Windows for Arm (still in beta) not work with your applications even if you specifically set them to run in compatibility mode within the windows environment)?

There are options and I’ll bet windows in azure will work for you. This isn’t running virtual box on Apple silicon btw.

Need for Windows is mostly for working with client files. Anything "in the cloud" is not as secure and keeping it local... even though such options can promise- and do deliver- high security benefits. Trusting a giant "hard drive in the sky" is trusting total strangers to keep their nose out of files stored there. Store them locally and you only have to trust those you definitely know with any access to that system/storage.

Similarly, Windows for ARM is great for apps that are able to reliably run on it. However, client apps are sometimes not exactly "up to date" and are iffy at best running on Windows ARM. Think like needing to run something only in Mac apps that never rolled on from 32-bit... or even a few things that didn't get upgraded from PowerPC (yes, there's still some of that kind of thing even in Appleland). How to deal with those situations? Be sure you have a old Mac(s) on hand to step back in time to when Macs COULD handle either. Same with Windows.

If you want maximum compatibility, you need real Windows, not Windows for ARM. Just as I have ZERO expectations that all current macOS apps will EVER be upgraded for native Apple Silicon (look back at Rosetta 1 to see the same sequence of events then), Windows has far more apps: a bigger chunk of them will never be updated for native Windows for ARM. So trying to tie Mac Windows compatibility to Windows for ARM or even the hypothetical Bootcamp 2 for Windows for ARM still misses the ideal target of broadest compatibility so that work with clients can always involve great confidence that what they need me to run can be run.

So yes, there ARE several options. And temporarily, hanging on to Intel Macs with bootcamp can cover this base too. But those days are numbered (I just had one such Mac conk). Perhaps we reach a day where Microsoft decides to sell Windows for ARM just like they sell regular Windows to anyone and everyone interested. And then perhaps the hints that Apple WOULD build a new Bootcamp for Silicon would occur too. But rather than wait & hope for both, the seemingly easier option is the one used BEFORE Macs adopted Intel CPUs.

All that offered, I DO think that for many people's needs Windows for ARM options could be 'good enough,' especially if the Windows apps they want/need to run are pretty mainstream such that they are likely already refined into code bases able to reliably run on ARM... or are already known to run just fine on ARM or X86 emulation on ARM. To defeat this 'good' option takes as little as only 1 app that you need to run for any reason that does NOT run reliably on ARM.

 

[DeepIn2U]

Need for Windows is mostly for working with client files. Anything "in the cloud" is not as secure and keeping it local... even though such options can promise- and do deliver- high security benefits. Trusting a giant "hard drive in the sky" is trusting total strangers to keep their nose out of files stored there. Store them locally and you only have to trust those you definitely know with any access to that system/storage.

Similarly, Windows for ARM is great for apps that are able to reliably run on it. However, client apps are sometimes not exactly "up to date" and are iffy at best running on Windows ARM. Think like needing to run something only in Mac apps that never rolled on from 32-bit... or even a few things that didn't get upgraded from PowerPC (yes, there's still some of that kind of thing even in Appleland). How to deal with those situations? Be sure you have a old Mac(s) on hand to step back in time to when Macs COULD handle either. Same with Windows.

If you want maximum compatibility, you need real Windows, not Windows for ARM. Just as I have ZERO expectations that all current macOS apps will EVER be upgraded for native Apple Silicon (look back at Rosetta 1 to see the same sequence of events then), Windows has far more apps: a bigger chunk of them will never be updated for native Windows for ARM. So trying to tie Mac Windows compatibility to Windows for ARM or even the hypothetical Bootcamp 2 for Windows for ARM still misses the ideal target of broadest compatibility so that work with clients can always involve great confidence that what they need me to run can be run.

So yes, there ARE several options. And temporarily, hanging on to Intel Macs with bootcamp can cover this base too. But those days are numbered (I just had one such Mac conk). Perhaps we reach a day where Microsoft decides to sell Windows for ARM just like they sell regular Windows to anyone and everyone interested. And then perhaps the hints that Apple WOULD build a new Bootcamp for Silicon would occur too. But rather than wait & hope for both, the seemingly easier option is the one used BEFORE Macs adopted Intel CPUs.

All that offered, I DO think that for many people's needs Windows for ARM options could be 'good enough,' especially if the Windows apps they want/need to run are pretty mainstream such that they are likely already refined into code bases able to reliably run on ARM... or are already known to run just fine on ARM or X86 emulation on ARM. To defeat this 'good' option takes as little as only 1 app that you need to run for any reason that does NOT run reliably on ARM.

Interesting you didn’t address suggestion of potentially of using as Azure Windows, have you looked into this? Regarding client security many corporations have moved on premises exchange to Azure or O365 from Financial to mining etc. many of which I’ve personally worked at.

I have a question regarding security of client files just how did you get those client files where they sent for the email and where it was the email unincripted did you get it in physical form VNSD card or US speed ask because if you got a email and it’s uninkripted there goes your concern a security

 

[HobeSoundDarryl]

I have looked into Azure. Again, don't love the option of "in the cloud" vs. simply buying my own Windows machine. If there was/is an "Azure"(type service) for macOS "in the cloud," I'd rather have a Mac on my desk too. Else, the same rationale for Windows that way should apply to Mac that way too. iOS Anyone? "Dumb" phone screen with iOS running in the cloud anyone (and just sending the dumb phone a screencast)?

Often, "we" can be quick to push remote Windows but can up to detest the idea of remote macOS. If there was such a service, would we want it instead of an actual Mac? For whatever reasoning we might have against that concept, the same probably applies to Windows too.

To get those client files, I generally pick them up on premises, while connected to their network, and/or get them onto external (tangible) storage while there OR they may bring them to me. On occasion, snail mail/FEDEX has been used (USB stick, etc). When security is not as important for select file exchanges, "the cloud" is used: Dropbox, Google, OneDrive, etc. But neither they nor I generally trust such options for everything.

 

[polyphenol]

My partner needs to use one piece of software which is only available on Windows. Further, it uses a dongle. And wants to use wifi to connect to a machine.

The idea of paying for an Azure instance, managing it, overcoming dongle issues, and maintaining it all is a ridiculous overhead. It is a hobby - not a major enterprise.

The best option for her would likely be if the company that sells the software offered an app instance (in Azure or wherever else), and she could just connect in and use the software. No maintenance. Run by the company, they could abandon dongles. And the wifi link could be achieved using almost any computer or tablet - if they set their minds to it.

 

[DeepIn2U]

Need for Windows is mostly for working with client files. Anything "in the cloud" is not as secure and keeping it local... even though such options can promise- and do deliver- high security benefits. Trusting a giant "hard drive in the sky" is trusting total strangers to keep their nose out of files stored there. Store them locally and you only have to trust those you definitely know with any access to that system/storage.

Similarly, Windows for ARM is great for apps that are able to reliably run on it. However, client apps are sometimes not exactly "up to date" and are iffy at best running on Windows ARM. Think like needing to run something only in Mac apps that never rolled on from 32-bit... or even a few things that didn't get upgraded from PowerPC (yes, there's still some of that kind of thing even in Appleland). How to deal with those situations? Be sure you have a old Mac(s) on hand to step back in time to when Macs COULD handle either. Same with Windows.

If you want maximum compatibility, you need real Windows, not Windows for ARM. Just as I have ZERO expectations that all current macOS apps will EVER be upgraded for native Apple Silicon (look back at Rosetta 1 to see the same sequence of events then), Windows has far more apps: a bigger chunk of them will never be updated for native Windows for ARM. So trying to tie Mac Windows compatibility to Windows for ARM or even the hypothetical Bootcamp 2 for Windows for ARM still misses the ideal target of broadest compatibility so that work with clients can always involve great confidence that what they need me to run can be run.

So yes, there ARE several options. And temporarily, hanging on to Intel Macs with bootcamp can cover this base too. But those days are numbered (I just had one such Mac conk). Perhaps we reach a day where Microsoft decides to sell Windows for ARM just like they sell regular Windows to anyone and everyone interested. And then perhaps the hints that Apple WOULD build a new Bootcamp for Silicon would occur too. But rather than wait & hope for both, the seemingly easier option is the one used BEFORE Macs adopted Intel CPUs.

All that offered, I DO think that for many people's needs Windows for ARM options could be 'good enough,' especially if the Windows apps they want/need to run are pretty mainstream such that they are likely already refined into code bases able to reliably run on ARM... or are already known to run just fine on ARM or X86 emulation on ARM. To defeat this 'good' option takes as little as only 1 app that you need to run for any reason that does NOT run reliably on ARM.

I still think, ahem know there are serious governmentally bound legalities that Microsoft MUST follow else huge stringent fines regarding corporate/business confidentiality and security of complete ‘total ownership’ of Azure based Windows hosted in the cloud. it’s right there on there site actually - resellers I’d completely ignore and go straight to the source.

That said I understand the need for a real machine for loads of reasons, and was just concerned it was not considered for you and a few others (not trying to sell anyone onto windows).

I think the only REAL way Microsoft would ever sell or give a real push for Windows on Arm builds a real effort is when Intel and AMD go full on ARM/RISC-V based CPU’s or combo CPU+GPU chips for laptops/desktops and drop the legacy code. Until then, which would force developers large (Microsoft, VMWare, AutoCAD, NVidia etc) and small (7Zip, WinDirStat) to compile software for ARM/RISC-V chips. Until the chip makers do it, most small developers will not take the time and re-create their software to do so. Force an environment change leads to developer’s change.

 

[HobeSoundDarryl]

That mentality of “force developers” seems logical but, in practice, is hit or miss… meaning the “force” works if there is enough tangible profit motivation to go to the trouble. Else, you get an abandonware scenario which is exactly what us Mac people have experienced at each tangible AppleOS change: deprecation of Rosetta 1, switches from 32bit to 64bit in both macOS and iOS and soon deprecation of Rosetta 2.

In reality, if some developers don’t anticipate enough profit for making a change, they will NOT bother… and any dependency on the “as is” version becomes a burden to users to hang on to an old machine or 2 (or 3) if any such software is important enough.

Thus, I maintain two old Macs capable of booting into Snow Leopard because it was the last OSX version capable of running Rosetta 1. A few abandonware apps never made the jump to Intel Mac native just as some won’t make the jump to Silicon and, should Microsoft ever try to “force” abandonment of X86, Windows for ARM (but much worse just because there are countless numbers of apps in that much larger market).

This mentality of “force” doesn’t really work. What does work is the money motivation to want to update the code. We’ll all learn this again soon when Apple pulls the plug on Rosetta 2. Even Apple knows “force” doesn’t work. That’s why we have Rosetta 2… for now.