ARM based MacBook Pro discussion [merged]

[skipjakk]

Stephen.R, you lost credibility with me when you mischaracterized what I said and then continued to argue against a position I never took(straw man argument). Taken all together, I find unofficial conversations from Intel developers, Intel officials, and forecasts from respected industry analysts credible. Trying to convince me otherwise is just a waste of your time.

 

[firewood]

Finally, given the differences iin thermal requirements, Apple might create custom silicon for their Mac lineup, without the thermal limitations of the A series chips...

Apple regularly designs new targeted custom silicon for watches and earbuds. If Apple introduces a macOS product with an ARM ISA application processor, it almost certain will have yet more custom silicon, instead of just being an iPad/iPhone A series chip. Thus potentially a lot more performant (IO bandwidth, etc.)

Who's gonna develop and change their apps to ARM-based Mac?

There are a lot more ARM-targeted app developers (iOS and Android) than pure Mac developers. The number of macOS apps might actually increase... maybe after a year to two. But tons of developers would buy in on day one, boosting sales ahead of the ramp up of more consumer macOS apps.

 

[high heaven]

There are a lot more ARM-targeted app developers (iOS and Android) than pure Mac developers. The number of macOS apps might actually increase... maybe after a year to two. But tons of developers would buy in on day one, boosting sales ahead of the ramp up of more consumer macOS apps.

Well, then that's a huge problem cause most professional software and programs exist in macOS.

What about Adobe software? FCPX? Logic Pro? Davinci Resolve? Nuke? Blender? Autodesk? etc? You see, no matter how many iOS app developers, there isn't many pro grade app available in iOS especially for serious works.

 

[DHagan4755]

Apple regularly designs new targeted custom silicon for watches and earbuds. If Apple introduces a macOS product with an ARM ISA application processor, it almost certain will have yet more custom silicon, instead of just being an iPad/iPhone A series chip. Thus potentially a lot more performant (IO bandwidth, etc.)

Right. Given how performant the A12X is in the iPad Pro, imagine Apple creating an ARM processor that uses active cooling — even if it's half the TDP of an 8-core i9 used in the 16" MacBook Pro. What will that do for ARM performance in a Mac notebook? Remains to be seen but the thoughts of what could be are tantalizing. I believe these ARM chips will be different enough from the ones used in iPads and iPhones that Apple will call it the X1 processor — or something along those lines.

 

[firewood]

What about Adobe software? FCPX? Logic Pro? Davinci Resolve? Nuke? Blender? Autodesk? etc? You see, no matter how many iOS app developers, there isn't many pro grade app available in iOS especially for serious works.

It's easy to play that game.

Think of all the Absolutely Must Have professional apps there were back in the day for Intel 32-bit, PowerPC, and 68k Macs. How many do you still use because they weren't replaced with something better?

Plus think of all the professionals and business executives who use iPhones and iPads for business applications, but don't use Macs. Advantage ARM for that serious work.

 

[danwells]

I have no problem seeing an ARM MacBook - running a mix of Apple internal apps (many if not most of which share codebase with iOS, and would be easy to port), apps from developers who do make the transition (again, if they have relatively similar iOS and Mac versions sharing a lot of code, it's not too hard), and Catalysted iPad apps. It's probably going to be locked to the Mac App store.

What I can't see is a MacBook Pro, which requires access to a wider range of apps, including some specific apps with massive legacy codebases - Adobe, I'm looking at you (and the iPad versions with a fraction of the functionality don't count). There are also many pro applications made by smaller companies who have little interest in iOS...

 

[high heaven]

It's easy to play that game.

Think of all the Absolutely Must Have professional apps there were back in the day for Intel 32-bit, PowerPC, and 68k Macs. How many do you still use because they weren't replaced with something better?

Plus think of all the professionals and business executives who use iPhones and iPads for business applications, but don't use Macs. Advantage ARM for that serious work.

We are talking about the present, not the past. If Apple fails to have pro-grade software for ARM-based Mac, it would be a failure. How come ARM-based Windows still not have pro-grade software?

 

[DHagan4755]

We are talking about the present, not the past. If Apple fails to have pro-grade software for ARM-based Mac, it would be a failure. How come ARM-based Windows still not have pro-grade software?

Many were expecting Final Cut Pro 10.5 last fall. Still nothing. Could this be partially the reason for the hold up?

 

[high heaven]

Many were expecting Final Cut Pro 10.5 last fall. Still nothing. Could this be partially the reason for the hold up?

Well, Final Cut Pro X got an update to have a Metal-based engine.

 

[firewood]

We are talking about the present, not the past. ... How come ARM-based Windows still not have pro-grade software?

Apple rarely cares about the past or even the present. Mostly only the future. That's why they can remove the absolutely necessary serial ports, floppy drives, and headphone jacks (etc.).

There are a ton more active iOS developers (and paying customers) than there are for any MS ARM product. Just compare the competing app store revenues.

 

[high heaven]

Apple rarely cares about the past or even the present. Mostly only the future. That's why they can remove the absolutely necessary serial ports, floppy drives, and headphone jacks (etc.).

There are a ton more active iOS developers (and paying customers) than there are for any MS ARM product. Just compare the competing app store revenues.

I said that's meaningless for professional users. No matter how many iOS apps there are, there arent any professional software that people can work with. Do you wanna make ARM based Mac without or only few pro apps?

 

[firewood]

Do you wanna make ARM based Mac without or only few pro apps?

History seems to indicate that Apple only needs a few (their own plus maybe a couple secret lead partners) to get the ball rolling.

 

[high heaven]

History seems to indicate that Apple only needs a few (their own plus maybe a couple secret lead partners) to get the ball rolling.

And this is a different situation that is not comparable to the past.

 

[Unregistered 4U]

it will be meaningless for pro users.

So, just taking that statement at face value... it would be meaningless to 10-20% of their users. The other 80% will be perfectly fine. Apple has dumped Pro users before in order to suit their own view of the future (64-bit only in Catalina cutting off 32 bit Pro apps, FCP7 users). Just saying... don’t be surprised if “being a pro” doesn’t hold a lot of reverence with Apple anymore

Not one of them has gone into the depth of discussion or thought that this thread did earlier

The main reason why this rumor has legs is because it’s from Intel. There have been rumors sources from random areas previously, but never has anyone ever been able to claim that the info came from Intel. And someone at Intel, of course, would be in a unique position to know with some certainly that “something” is happening while still having no idea what is happening.

which requires access to a wider range of apps, including some specific apps with massive legacy codebases

Won’t folks that need access to those massive legacy codebases just keep the computer they have? There are people currently working on systems that are 5-8 years old, still doing professional work on them. Today’s MBP’s will last those folks well into the future. In my mind, any future MBP just needs to run Apple’s Pro apps. Adobe could decide to recompile their apps, or maybe not. They’ve got a subscription model now, so they don’t care what platform you’re running it on, just as long as you’re running it.

 

[Woochoo]

To all the "you can't do professional/scientific stuff in ARM" guys: there are ARM-based supercomputers, and supercomputers are precisely used for that kind of stuff.

ARM is just like x86, they are both RISC based and the only difference is x86 has a decoder for complex instructions that translates into reduced instructions (actually it only has that from CISC, the rest is RISC. The pipelines and the rest are really similar. Heck, you can even compile the same binaries for x86 and ARM with their respective toolchains.

The problem is people has a mindset of ARM being low power because that's what they've been used for till now: mobiles with limited power constrictions (small battery, passive dissipation on a small enclosure). But there isn't any architectural limitation that stops them going for high clockspeed and high single core/multicore performance, for instance an old 28nm A9 already hit 3,1GHz and it wasn't even designed for that.

So if Apple is already delivering laptop performance ARM chips in their horribly thermal designed iPhones and iPads (compared to active thermal designs with fans and heatsinks), why couldn't they deliver something much better specially designed for high performance in a laptop? An iPad Pro already beats a 2018 15" MBP in rendering an edited video, I'd say that's not a "misleading synthetic benchmark" (see Geekbench or any other crossplatform)

Now on the apps part like Blender and all that stuff that someone was mentioning before: they can be recompiled with the right toolchain into ARM, they already have the code done. Maybe they will have to change some optimizations like AVX/SSE instructions to NEON, but not much more. Also if Apple implements Catalyst in the right way both x86 and ARM should be able to run in the same OS and processor without not much overheat (unlike that infamous Windows 10 running in a SnapDragon 835 with tons of layers translating the instructions making it slow).

 

[bo-waleed]

Isn't 2020 the year apple supposed to release them ?

Instead we are seeing rumors for update just for Intel last CPUs,

I want Apple to move to ARM as soon as possible to make App Devolpers take their time and focus on ARM and utilize it's potential.

And hopefully make Apps that emulates old Intel based Apps.

 

[theluggage]

Isn't 2020 the year apple supposed to release them ?

Only according to a sketchy hearsay rumour about something that "Intel expected" which could be:
(a) Complete hogwash
(b) A snippet from Intel's "due dilligence" worst-case scenario risk register.
(c) Arguably fulfilled if Apple updates the iMac with a T2 chip, in which case all Macs will have an ARM chip - just not as the main CPU. OK, that's a bit of a stretch, although...**

Instead we are seeing rumors for update just for Intel last CPUs,

If an ARM transition did happen, it couldn't happen overnight, and they'd need to be selling credible Intel Macs alongside ARM NewMacs for a few years. As I said below, I think some people are over-blowing the difficulties, but I think that's partly because they're imagining being forced to switch overnight. What they really stuffed up with the 2016 MBP, the 2018 Mac Mini and the 2019 Mac Pro was letting their predecessors get so out of date that people were desperate to upgrade, then introduced drastic functionality and pricing changes - forcing people to re-think their workflow overnight (or stick with their old hardware).

I want Apple to move to ARM as soon as possible to make App Devolpers take their time and focus on ARM and utilize it's potential.

In 2020 the vast majority of developers writing new code shouldn't need to know or care what CPU their code is running on and use hardware abstraction frameworks wherever possible (Metal, Accelerate, Core Audio etc.) even most hardware drivers can and should be written in compiled languages and work through OS calls. Including specific CPU code without very, very good reason (of course, there are always exceptions) - is just bad design, and might not take advantage of even future hardware in Intel Macs (e.g. the Afterburner cards for the new Mac Pro).

...bear in mind that testing iOS apps in XCode already works by compiling the code for x86 and running it in an iOS-for-x86 sandbox.

In a way, Apple have been preparing for ARM, in the sense that all of the App Store guidelines, dumping 32-bit, promoting Metal have been pushing people towards CPU-independent coding.

The big exception, of course, is established applications with legacy code dating back to Stonehenge (or, worse, 16-bit DOS/Windows). However, I have no idea how much of that has survived the transitions from PPC to x86-32 and x86-32 to x86-64 - and neither does anybody else unless they're actually working on such a codebase. The recent switch to 64-bit only will have culled a lot of ancient code and abandonware anyway.

Still, yes, "feasible" doesn't mean "trivial" and the first step would have to be a development system to allow applications to be re-compiled and tested. That's actually what happened with x86 - Apple produced what was basically a Hackintosh as a developers-only system about 6 months before launching the MacBook Pro (...and maybe key developers got it under NDA before that).

And hopefully make Apps that emulates old Intel based Apps.

...that could be the problem. Of course it is technically possible, but there might be some IP issues with emulating the x86-64 instruction set (I believe the emulator in ARM Windows is 32-bit only).

The fundamental problem is that the raw processing speed of an ARM core is unlikely to be much faster than the Intel core it is replacing. Previous transitions (6502 to 68k, 68k to PPC, PPC to x86) have all come with a substantial boost in raw speed to offset the overhead of emulation.

The advantage of ARM is more about being able to cram on more cores and more on-chip accelerators and GPUs which will only be available to applications using standard MacOS frameworks (which will be strongly correlated to being relatively easy to port to ARM). So I'm not sure that emulation is going to be as key as it was in PPC-x86 or 68k to PPC.

Thinking about it, what if a new Mac couldn't run MS Word on day 1? In 2005, that would have been suicide. In 2020, with so many alternatives (like Pages, platform-independent web apps, or LibreOffice which already supports ARM Linux) it might be merely "courageous".

** Now, this is just speculation, but how far could Apple extend the T-series chips to include things like GPUs and USB4 (i.e. Thunderbolt in all but name) - effectively making them into A-series-like systems-on-a-chip, replacing a load of Intel motherboard & Thunderbolt chippery but keeping the Intel (or AMD) processor...?

 

[DHagan4755]

Look, the rumors have been swirling for many, many years. There is smoke. Some of the factors in switching to ARM have been trivialized, while some of it has been foreshadowed by software decisions Apple's made in recent years.

One one level, if an ARM MacBook Air ran for many more hours on a battery, has better performance than an Intel Y-series chip, and could be offered at a lower price point, it would make sense to start transitioning to ARM. On another level, some applications that need to be re-written for ARM may not be as easily accomplished by clicking a compile for ARM processors check box in Xcode. This is where rift is.

I would wager that Apple's got Macs running on ARM in their labs right now. And probably have had them for years at this point. There was this anonymous posting at Slashdot from 2017 (3 years ago today) that was an interesting read... https://hardware.slashdot.org/comments.pl?sid=10191963&cid=53786433

 

[Stephen.R]

Look, the rumors have been swirling for many, many years. There is smoke.

there have been rumours of big foot, the Loch Ness monster and the abominable snow man for decades.

 

[DHagan4755]

there have been rumours of big foot, the Loch Ness monster and the abominable snow man for decades.

Haha yes there has. But there is more than enough reason to believe ARM Macs are possible than Big Foot, Loch Ness, and the Abominable Snow Man.

 

[theluggage]

I would wager that Apple's got Macs running on ARM in their labs right now.

I wouldn't take your money - it would be almost negligent of Apple not to have prototype ARM and AMD systems to hedge against future problems with Intel (and act as a bargaining chip) and if there isn't at least an unofficial "skunkworks" project on the go to get MacOS running on an iPad Pro then there's something very wrong with the culture of Apple...

Of course, plausible speculation is still speculation, and how functional or marketable such a system might be is another matter.

But there is more than enough reason to believe ARM Macs are possible than Big Foot, Loch Ness, and the Abominable Snow Man.

MacOS 11.0 : Bigfoot
MacOS 11.1 : Nessie
MacOS 11.2 : Yeti
MacOS 11.3 : Snow Yeti

...you heard it here first. Apple caught the actual creatures in 1999 so they could get an exclusive on the names, since when they've been kept in a basement at Cupertino next to the ARM, Transputer, Alpha, Itanium, Cell and RISC-V Macs... because that's the only place they could have been for the last 20 years and not be spotted by the 130% of the population who now walk around with a digital camera permanently fused to their hand or on their car dash.

Seriously, though: even Microsoft has Windows on ARM and they're far more shackled by legacy software than Apple ever will be - they're just dropping support for a 10-year old version of Windows after a 5 year transition period to an OS that is still mostly compatible with old software - and the wailing and gnashing of teeth is mainstream, non-tech news. One of the big differences between Mac and Windows is that Apple can and does regularly force change on their user base and kill off legacy functionality.

 

[pshufd]

I've heard this rumor since 2008. Software companies will leak porting projects before it happens so best just to ignore the idea and move on with life.

 

[danwells]

There are at least three BIG applications or sets thereof that cause porting issues... One (actually a set of applications that work similarly) simply can't be ported. The others could be, but both caused HUGE issues at the x86-PPC and PPC-Intel transitions, will probably take a long time to port.

1.) Parallels, VMWare, etc. These simply won't port, because they rely in the PC that underlies a Mac to work. There were predecessors for the PowerPC (was one called VirtualPC?),but they caused 50-80% performance hits because they were actually emulating a different processor, rather than merely running a virtualized instance of the processor they have. If Macs go ARM, it's right back to emulation and 50%+ performance hits to run Windows software (except for rare Windows-on-ARM software).

2.) Microsoft Office. Microsoft doesn't have it ported to ARM on their own OS on their own hardware - the Surface Edge is running big chunks of Office in emulation (very slowly). It was very slow to move on two previous processor transitions. It has a huge, fusty old legacy codebase (supposedly including large chunks of Windows). What makes anybody think it'll move quickly this time? We may very well see Catalysted iPad Office instead of full desktop Office (and MS may or may not ever offer the full version for ARM Mac).

3.) Adobe Creative Cloud. A huge, fusty old legacy codebase again (I've heard that there's x86 Assembly code in some parts of Photoshop, although I'm not sure if that's true). Again, a massive pain at the two previous transitions. iPad versions of many of the components exist, and are missing critical functionality. Will Adobe bother to port anything other than the iPad versions? In the case of Lightroom, they're actively trying to force photographers over to a new cloud-centric version based on the iPad codebase, rather than the more functional desktop version based on an older codebase...

 

[turbineseaplane]

@danwells

Adobe is getting onboard with better and better ARM iOS software...

...and Apple doesn’t care a lick about MS or Virtualization issues you raised.

Sorry. Just being honest.

I agree with you that they are concerns but no part of me believes Apple does.

 

[pshufd]

Machine-Code translation is an option for running on incompatible architectures but it isn't 100%.

Assembler code is used in many large programs where performance of a small segment of code that is used a lot can benefit. I think that you tend to have more intrinsics support in compilers today but you may lose some decent performance emulating it on an incompatible architecture.

One other area, and it's a big one, for certain applications. Intel arguably makes the highest-performing compilers in the x86 world, so much so that AMD uses them in their benchmarks. They also provide libraries of high-performance software libraries in multimedia process, math processing and engineering applications. They're not going to port these for ARM.