[Bloomberg] Apple starting process to dump Intel in Macs

[jpiniero]

The context was about Intel abandoning support for 16/32bit code with next generation CPUs.

That's fair, but the Rapids at this point are rumored to be server only. The question then becomes much of an effort is Intel going to put into improving performance of the Lakes with their focus on the Rapids.

 

[dark zero]

32-bit Windows applications (games, especially), used Win16 installers. And when most apps moved to 64-bit, they still now use 32-bit installers.

If you eliminated 32-bit mode, you would lose access to wide swathes of software, that were themselves, "64-bit".

So is Idiot programming on a massive scale?
My bad with that....

 

[PeterScott]

So is Idiot programming on a massive scale?
My bad with that....

The 64bit x86 chips fully support 32 bit code with no significant overhead, so there is really no need to stay away from 32 bit code.

There is no point moving simple software like installers, to 64 bit, when they gain no benefit from it.

Smart programmers don't waste time on inconsequential things.

Given the history of backward compatibility in x86 and Windows, there is no real expectation that 32 bit support is going away in the foreseeable future.

 

[ksec]

I think there are possibilities of Wintel providing backward compatibility in software mode. 32bit x86 is still available as emulation.

Given the other thread about Intel's leaks, it seems strange Apple decide to dump them now. ( Or those "leaks" are purely false )

 

[PeterScott]

I think there are possibilities of Wintel providing backward compatibility in software mode. 32bit x86 is still available as emulation.

Given the other thread about Intel's leaks, it seems strange Apple decide to dump them now. ( Or those "leaks" are purely false )

I think all of these stories are just insubstantial rumor at this point.

Many of them are the "wouldn't be neat if X happened" things that have already crossed many peoples minds. That makes them good rumor fodder.

But many also have downsides that get more significant, the deeper you dig into it them. Which makes them more questionable.

These things could happen, but I don't feel very convinced any of these rumors have much substance.

 

[Eug]

It’s pretty much a certainty that Apple has macOS running on ARM. It’s also likely they are developing fully spec’d ARM machines that run macOS. Probably they intend on releasing this at some point but whether or not that happens and when are not always going to be certain.

For example Apple has made fully functional cellular MacBook Pro prototypes in the past and one actually made it into the wild but it didn’t end up getting mass produced.

 

[PeterScott]

There is little doubt that Apple keeps their OS source code CPU architecture agnostic, and compile it for both targets.

iOS is based on MacOS, and they cross pollinate bag and forth. It would be shocking if they didn't maintain for both targets.

But that doesn't mean they are getting ready for a switch or even that they are hedging for the possibility, just that having products in a different architectures that come from the same original source, they are making it easier to continue to move code between them.

So the existence of MacOS that can compile for ARM really doesn't signal anything about a switch.

I am sure they have iOS that will compile on x86, and no one would think that means they are switching iPhone to x86.

My biggest bit of skepticism on this is how messy, expensive and work intensive, this would be, contrasted against how little effort/attention Apple has shown Macs in the last decade. Macs are left to languish running old parts for years (how old are Mac Pro and Mac mini?).

Macs are a small business to Apple, and they treat it as an afterthought.

That they are suddenly go to spend big, on money/resource/time on a transitions of somewhat questionable benefit, really doesn't make a lot of sense to me.

I just don't believe Apple cares that much about Macs anymore to make that kind of effort.

 

[Eug]

There is little doubt that Apple keeps their OS source code CPU architecture agnostic, and compile it for both targets.

iOS is based on MacOS, and they cross pollinate bag and forth. It would be shocking if they didn't maintain for both targets.

But that doesn't mean they are getting ready for a switch or even that they are hedging for the possibility, just that having products in a different architectures that come from the same original source, they are making it easier to continue to move code between them.

So the existence of MacOS that can compile for ARM really doesn't signal anything about a switch.

I am sure they have iOS that will compile on x86, and no one would think that means they are switching iPhone to x86.

My biggest bit of skepticism on this is how messy, expensive and work intensive, this would be, contrasted against how little effort/attention Apple has shown Macs in the last decade. Macs are left to languish running old parts for years (how old are Mac Pro and Mac mini?).

Macs are a small business to Apple, and they treat it as an afterthought.

That they are suddenly go to spend big, on money/resource/time on a transitions of somewhat questionable benefit, really doesn't make a lot of sense to me.

I just don't believe Apple cares that much about Macs anymore to make that kind of effort.

This is not about creating 25 different chip SKUs. This is about scaling up their designs to encompass more than just iPhones and iPads. Given that they sell tens of millions of Macs every year, it makes perfect sense.

 

[PeterScott]

This is not about creating 25 different chip SKUs. This is about scaling up their designs to encompass more than just iPhones and iPads. Given that they sell tens of millions of Macs every year, it makes perfect sense.

I think you are drastically underestimating both the effort involved, and Apple's Mac neglect.

 

[jpiniero]

I think you are drastically underestimating both the effort involved, and Apple's Mac neglect.

Apple hasn't neglected the Mac at all... just the desktops. The Macbook is having issues but that's only because of the disaster that is Intel's 10 nm. And then there's the lack of LPDDR4....

 

[ksec]

I wonder if a Dual ISA , in a clean and elegant way is possible for software developers.

Say Apple has their SoC as main chip and x86 as co-processor for a few Mac models.

 

[PeterScott]

I wonder if a Dual ISA , in a clean and elegant way is possible for software developers.

Say Apple has their SoC as main chip and x86 as co-processor for a few Mac models.

Frankenstein monster, not worth building.

 

[Eug]

I think you are drastically underestimating both the effort involved, and Apple's Mac neglect.

Translated: “It’s too hard / expensive. Apple can’t / won’t do it.”

Notably, this is the argument people made about the idea of Apple making its own custom ARM SoC instead of buying off the shelf. This is also the same argument about the idea of Apple making its own custom GPU from the ground up.

 

[ksec]

Frankenstein monster, not worth building.

Because I just dont see how ARM plays into it. Yes we all know it is rumours. But Mark Gurman has a very reliable record on any Apple Software related news ( His hardware prediction is awful, KGI's Ming is much better in this regard ). So I guess this could be real.

Translated: “It’s too hard / expensive. Apple can’t / won’t do it.”

Notably, this is the argument people made about the idea of Apple making its own custom ARM SoC instead of buying off the shelf. This is also the same argument about the idea of Apple making its own custom GPU from the ground up.

It is not in anywhere similar. It absolutely make sense for Apple to made their own SoC. And for GPU i only question whether it was feasible in terms of GPU patents. (Turns out it was ) Apple Ship 250M iPhone and close to 300M iOS devices, that is more then the total PC market with Intel and AMD combined. And doing it using only 2 Chips per year. The AX variants as pipe cleaner for New Node on iPad Pro, and new A series for iPhone. Compared to a few dozens of variation on PC market from Intel.

Apple ships roughly 25M unit of Mac each year, 5M of those are Mac Mini, iMac, Mac Pro. You could have only a few variation of A10X for the 20M Macbooks. But you need a lot of different CPU for the 5M Desktop Mac market with 45W to 240W CPU. Designing a 240W TDP CPU is completely different to a 5W or 15W SoC. You will use a new 7nm Node for HP, new verification tools, new QA. Two / Quad Memory Channel. ECC memory support and testing. The 20M of Macbook CPU could share many of the uses with iPhone and iPad, but not the same for Mac CPU. That is why it doesn't make any sense financially speaking.

It will be messy, with no immediate upside, and expensive. That is why PeterScott kept saying it is not worth it. And why I suggest a Dual ISA or using Qualcomm Centriq for higher end Mac, or Apple using their own CPU in Datacenter. Because so far none of these make any sense at all.

 

[Eug]

Because I just dont see how ARM plays into it. Yes we all know it is rumours. But Mark Gurman has a very reliable record on any Apple Software related news ( His hardware prediction is awful, KGI's Ming is much better in this regard ). So I guess this could be real.



It is not in anywhere similar. It absolutely make sense for Apple to made their own SoC. And for GPU i only question whether it was feasible in terms of GPU patents. (Turns out it was ) Apple Ship 250M iPhone and close to 300M iOS devices, that is more then the total PC market with Intel and AMD combined. And doing it using only 2 Chips per year. The AX variants as pipe cleaner for New Node on iPad Pro, and new A series for iPhone. Compared to a few dozens of variation on PC market from Intel.

Apple ships roughly 25M unit of Mac each year, 5M of those are Mac Mini, iMac, Mac Pro. You could have only a few variation of A10X for the 20M Macbooks. But you need a lot of different CPU for the 5M Desktop Mac market with 45W to 240W CPU. Designing a 240W TDP CPU is completely different to a 5W or 15W SoC. You will use a new 7nm Node for HP, new verification tools, new QA. Two / Quad Memory Channel. ECC memory support and testing. The 20M of Macbook CPU could share many of the uses with iPhone and iPad, but not the same for Mac CPU. That is why it doesn't make any sense financially speaking.

It will be messy, with no immediate upside, and expensive. That is why PeterScott kept saying it is not worth it. And why I suggest a Dual ISA or using Qualcomm Centriq for higher end Mac, or Apple using their own CPU in Datacenter. Because so far none of these make any sense at all.

Again, there is no requirement for Apple to design a ton of different CPUs to encompass everything from the 12” MacBook to the MacPro. All they need is something that can cover the main consumer machines initially.

BTW, Apple only sold 11 million iPhones in 2008. That’s after Apple bought P.A. Semi for their CPU design talent, after they sold 1.4 million iPhones in 2007. Their first in-house CPU came out in 2010.

 

[oak8292]

Apple ships roughly 25M unit of Mac each year, 5M of those are Mac Mini, iMac, Mac Pro. You could have only a few variation of A10X for the 20M Macbooks. But you need a lot of different CPU for the 5M Desktop Mac market with 45W to 240W CPU. Designing a 240W TDP CPU is completely different to a 5W or 15W SoC. You will use a new 7nm Node for HP, new verification tools, new QA. Two / Quad Memory Channel. ECC memory support and testing. The 20M of Macbook CPU could share many of the uses with iPhone and iPad, but not the same for Mac CPU. That is why it doesn't make any sense financially speaking.

ksec

In general I agree with you sentiment on the probability of Apple producing their own CPU, however I will play devil's advocate. AMD is headed down the path that many see as the next step in keeping Moore's 'Law' alive with EPYC. The 'chiplet' concept is being used to reduce cost on a server class processor. This concept could migrate to lower cost processors to improve volumes. If Apple used a 10-20 watt 4 core 'chiplet' for their Mac line up they might be able to make it work. The chiplets would have a higher volume, better yield and provide flexibility in making 4, 16 or even 24 core processors. The 2.5 D or 3 D technology is being developed by all the foundries and the packaging companies.

In general I think there are very good reasons to keep iOS and Mac OS on separate paths for awhile longer that has never stopped Apple. Apple is often making transitions to the next thing before many of their customers are ready.

 

[ehume]

I know it is late to chime in on the compatibility issue, but I work, see . . .
We keep an ancient Dell in the basement because our Windows machines have not been able to run 1990's games sine WinXP. And I just installed PaintShopPro 7 (an old 32-bit app) on my wife's Win10 replacement machine (the failed case-mounted switch on a 2008 Dell was unreachable). And we are not remarkable.

In my experience people are holding on to their old machines a long time. Oh, they upgrade their operating systems when pressed, but they let theirapps stay. Even mildly advanced users who can replace their HD's with SSD's find that the performance boost so gained breathes new life into their old machines. So Apple might be able to leave old users behind; not so in the PC world.

 

[PeterScott]

I know it is late to chime in on the compatibility issue, but I work, see . . .
We keep an ancient Dell in the basement because our Windows machines have not been able to run 1990's games sine WinXP. And I just installed PaintShopPro 7 (an old 32-bit app) on my wife's Win10 replacement machine (the failed case-mounted switch on a 2008 Dell was unreachable). And we are not remarkable.

I run games from the 1990's on my Win7-64 machine (Total Annihilation, Baldurs Gate).

Microsoft has done an Amazing job with backward compatibility. You can probably find a way to get most old games running.

I would hate to have to keep an ancient machine around to run old software. What happens when that machine inevitably fails?

 

[ksec]

Again, there is no requirement for Apple to design a ton of different CPUs to encompass everything from the 12” MacBook to the MacPro. All they need is something that can cover the main consumer machines initially.

How would that work? iMac sticks to x86? You have some software which works and some dont? Or you expect Rosetta really to be magic and "just works". Because in reality Rosetta is for from perfect. Universal binary? It is still lots of work for pro Apps, I.e it isn't as simple as apple said just recompile.

In general I agree with you sentiment on the probability of Apple producing their own CPU, however I will play devil's advocate. AMD is headed down the path that many see as the next step in keeping Moore's 'Law' alive with EPYC. The 'chiplet' concept is being used to reduce cost on a server class processor. This concept could migrate to lower cost processors to improve volumes. If Apple used a 10-20 watt 4 core 'chiplet' for their Mac line up they might be able to make it work. The chiplets would have a higher volume, better yield and provide flexibility in making 4, 16 or even 24 core processors. The 2.5 D or 3 D technology is being developed by all the foundries and the packaging companies.

That is certainly an interesting possibilities! Can't think of a reason argue against it!

 

[Eug]

How would that work? iMac sticks to x86? You have some software which works and some dont? Or you expect Rosetta really to be magic and "just works". Because in reality Rosetta is for from perfect. Universal binary? It is still lots of work for pro Apps, I.e it isn't as simple as apple said just recompile.

Apple would compile all of their important software to be cross platform, or cross-architecture, as it were. Not all third parties would but it wouldn't actually matter as much for the lower end machines, at least early on. Furthermore, Apple is apparently already beginning to address this, in advance, with their supposed Project Marzipan.

While it would not necessarily be an easy task, it Apple gets this right, it would be easier than before.

Also, I will admit I am not a developer, but this article about UIKit and UXKit was an interesting read for me:

https://medium.com/@guilhermerambo/why-uikit-for-macos-is-important-ff4e74a82cf0

UXKit for Mac is a private framework for Apple that was used to develop Photos, but which is basically a mirror to UIKit for iOS. A lot of developers complain about macOS' AppKit, and wish development on the Mac was more like UIKit on iOS. UXKit begins to fulfill that wish. The main problem is that UXKit is not a public framework... yet.

What about MS Office? Well, even though it would still be on AppKit, Apple would pay Microsoft big bux to get MS Office working well on macOS ARM. Same thing for Adobe Creative Suite.

 

[beginner99]

This is also the same argument about the idea of Apple making its own custom GPU from the ground up.

This is mostly a patent issue. Not knowing much on what exactly is going on and from whom they are licensing my bet is they are either planning to buy imagination or just outright ignore their patents knowingly imagination doesn't have the money for a decade long legal battle.

 

[Nothingness]

This is mostly a patent issue. Not knowing much on what exactly is going on and from whom they are licensing my bet is they are either planning to buy imagination or just outright ignore their patents knowingly imagination doesn't have the money for a decade long legal battle.

I don't think Apple would be stupid enough to ignore patents issue: Imagination could be acquired by some patent troll company just to attack Apple. And alas there are a lot of patent troll companies with enough money to damage the industry.

 

[ksec]

This is mostly a patent issue. Not knowing much on what exactly is going on and from whom they are licensing my bet is they are either planning to buy imagination or just outright ignore their patents knowingly imagination doesn't have the money for a decade long legal battle.

Actually that is one of the biggest mystery that we may never know. My bet is that they license all GPU related patents from ARM as well. Given how cheap Mali as an GPU implementation is, only the patents will likely be much cheaper.

 

[Nothingness]

Given how cheap Mali as an GPU implementation is, only the patents will likely be much cheaper.

Mali is cheap because ARM likely makes a discount when picking a GPU license along with a CPU license

 

[jpiniero]

https://developer.apple.com/macos/whats-new/

Apple is depreciating OpenGL in the next version of macOS, which means the only supported Graphics API will be Metal. I imagine it helps with the ARM transition but it does show how little they care about backwards compatibility.