Arm Powered Mac Must Be Powerful, But Why/When Top Dogs Would Join The Party?

[alxhxxa]

I wouldn't doubt about the abilities of the Apple Chip design team, I believe that ARM-Macs would be very powerful, as good as any equivalent Intel-Mac at the very least, doesn't matter what hardware specs on paper are. What I worry is, while 85% of the market is dominated by the x86 base machine, why would all those big software companies want to join the party ? I'm not a software engineer, but I don't believe that's a easy job at all. And even if they do, how many years would it takes for Mac users to be benefited, practically speaking ?

Until then, an ARM-Mac could only be a "very powerful" computer for Apple-Apps and "ultra-powerful" computer for iPhone/iPad apps.

I love my iPhone and iPad Air, but I just don't believe that they can rock the x86 world, it's just too big......

Unless the ARM-Mac is nearly as good as a Quantum Computer, that'll do ?

 

[leman]

What I worry is, while 85% of the market is dominated by the x86 base machine, why would all those big software companies want to join the party ?

I don’t understand the question. Developing for the ARM Mac is not any different to developing for the x86 Mac. If your logic would apply, we wouldn’t have any Mac software, because the software developers would focus only on Windows. And yet there is plenty Mac-only software, which means it must be profitable enough.

Also, while Mac is only a small fragment of the market, it’s a substantial part of the “premium” market. Most x86 machines out there are cheap, low performance office setups. When you take the users who do more serious work, you’ll find quite a few Macs there. Gamers excluded of courses, but fast gaming machines are a minority of all shipped PCs.
[automerge]1593974140[/automerge]

I'm not a software engineer, but I don't believe that's a easy job at all. And even if they do, how many years would it takes for Mac users to be benefited, practically speaking ?

For majority of apps, it takes one click to get the a working ARM version. Of course, you need to test etc... in the worst case you have to reimplement some platform specific code. But it’s less effort than you assume.

 

[Jorbanead]

Apple is one of the few companies that can get away with this. Apple generally sets new trends in the industry, and many are predicting that there is going to be a new surge in ARM chips for the PC market (though it will take maybe another decade for the rest of the market to catch-up). There is already a version of Windows 10 that can run off of ARM, though its not very great at the moment. I fully expect to see HP, Dell, and others to realize an ARM laptop within the next decade. The power efficiency is just too good to pass up, and Apple is going to do a lot of the groundwork getting the industry onboard. Once the major players like Microsoft and Adobe get fully onboard, many others will follow.

That being said, Apple has done a lot of groundwork already over the last 5+ years. We already know that big players in the industry are already working on ARM versions of their platforms. Im not a developer, but I did watch all of the WWDC videos, and one thing Craig kept stressing was how easy it was to create an ARM version of any x86 software. He said even for demanding apps like Photoshop, they only had one person from Adobe come over (because they wanted to keep this super secret). I believe he said it took them like 2 days to create an ARM version of photoshop - with just one guy. So assuming this is true, it sounds like it won't be too painful for developers to transition.

Knowing Apple, they have also already been working with AVID and they probably already have ARM versions of Media Composer and Pro Tools running. But knowing AVID they won't release it until they've tested it a million times over. I would guess by the time an ARM Mac Pro comes out, all of the major players will have all of their software running on ARM.

 

[TiggrToo]

but I just don't believe that they can rock the x86 world, it's just too big

Unsure why you think this is the objective.

It's not. As far as Apple is concerned the rest of the x86/x64 world can go pound sand.

This is simply to reduce reliance on an extremely unreliable vendor whose now got a record or producing chips with since extremely horrible security holes.

Given Apple's clear victories when the A series SOC's, it makes total sense that Apple should seize this to bring everything in house so they can recreate the magic of yesteryear and produce the laptops of tomorrow.

Right now folk complain about Mac performance and say that Apple has lost its way.

Here's the response. They were listening and they agreed.

 

[leman]

I fully expect to see HP, Dell, and others to realize an ARM laptop within the next decade. The power efficiency is just too good to pass up, and Apple is going to do a lot of the groundwork getting the industry onboard. Once the major players like Microsoft and Adobe get fully onboard, many others will follow.

ARM is not necessarily inherently more power efficient, it’s that it was traditionally used for CPUs with low power consumption. Apple CPUs are not this fast and power efficient because they use ARM ISA, but simply because Apple has designed one hell of a CPU. The problem for other vendors is that they don’t have access to ARM CPUs of this caliber and it will probably take a while until such CPUs appear.

 

[jdb8167]

ARM is not necessarily inherently more power efficient, it’s that it was traditionally used for CPUs with low power consumption. Apple CPUs are not this fast and power efficient because they use ARM ISA, but simply because Apple has designed one hell of a CPU. The problem for other vendors is that they don’t have access to ARM CPUs of this caliber and it will probably take a while until such CPUs appear.

They wouldn’t design their own SoC but use the latest design from ARM itself. ARM has a couple of high end designs that are quite a bit better than Qualcomm’s latest which is why Qualcomm is also going to use them.

 

[ruslan120]

ARM is not necessarily inherently more power efficient, it’s that it was traditionally used for CPUs with low power consumption. Apple CPUs are not this fast and power efficient because they use ARM ISA, but simply because Apple has designed one hell of a CPU. The problem for other vendors is that they don’t have access to ARM CPUs of this caliber and it will probably take a while until such CPUs appear.

A lot less instruction translating, single x86 instructions like “copy string from memory location” require dozens of ARM instructions, ARM generally saves on instruction translation logic (done at compile time) at the cost of larger instruction memories required for the same work done.
[automerge]1593978606[/automerge]

They wouldn’t design their own SoC but use the latest design from ARM itself. ARM has a couple of high end designs that are quite a bit better than Qualcomm’s latest which is why Qualcomm is also going to use them.

It’s a bit unfortunate* that Apple’s custom ISA extensions to ARM make devices like the Raspberry Pi incompatible as a Hackintosh. Apple uses the ARM core with a plethora of its own (proprietary) additions.

* Awesome for the performance of actual Mac / iOS devices, just unfortunate from a compatibility perspective, would’ve been a cool project
[automerge]1593978714[/automerge]
Would like to echo the thought that actual development isn’t hard. Code written in swift will still be compatible since a ”compiler” takes swift and then translates it into either x86 or ARM, most developers don’t actually touch x86 or ARM. Java will still be slow. Some libraries might need to be re-written (low level code), that’s pretty much it.

 

[leman]

A lot less instruction translating, single x86 instructions like “copy string from memory location” require dozens of ARM instructions, ARM generally saves on instruction translation logic (done at compile time) at the cost of larger instruction memories required for the same work done.

“Dozens” might be a slight exaggeration. Yes, ARM uses separate instructions fir memory loss/store where Intel can combine this into the main instruction. But that’s about it. Both ARM and Intel will decode instructions into multiple micro-ops that are the “real” commands that CPUs understand.

Where ARM can indeed save some complexity compared to x86: decode logic unit (ARM instructions are fixed length and have more straightforward format) and no need to support legacy crap. x86 CPUs have to support all the obsolete stuff like 32bit ISA, x87 fp operations and so on. But it’s not like ARM. plus need to do less work here.

 

[Joelist]

Apple transitioning to Apple Silicon is a different proposition than, say, Dell putting out Windows laptops using ARM.

For one thing, Apple Silicon probably should be termed "ARM-ish", as while they do use the basic instruction set (per their Architecture license) they have added quite a bit to it and the cores and controllers are a totally Apple design (and a superb one). Also as they are the providers of the OS and all drivers these are precisely written to the architecture.

Meanwhile, the Dell in our example has to buy their SOC from someone else and it is more or less straight ARM, which means that it struggles under some scenarios which are more common to full PCs than to devices. Also it is using an OS and drivers that are nowhere clear as closely written to the hardware - and I like Windows 10 nowadays. It started out rough but matured.

So the Dell already is suboptimal. Now add in that there is a universe of Windows applications that are x86 - so someone has to provide an emulator or else this machine is dead out of the gate.

None of this is to say that only Apple can do this - just that anyone else will have some hoops to jump before they can realistically try.

 

[theluggage]

What I worry is, while 85% of the market is dominated by the x86 base machine, why would all those big software companies want to join the party ? I'm not a software engineer, but I don't believe that's a easy job at all.

You're over-estimating the amount of work required to support different CPUs under the same operating system - which is tiny compared to the effort needed to port applications between operating systems, and Apple has made it easy by providing the compiler technology, universal binaries etc. If a developer already thinks it is worth supporting the Mac then they will make the change - any that drop out because of ARM were probably on the verge of giving up anyway. The recent switch from 32-bit to 64-bit was probably a bigger deal.

Of course, there will be a few exceptions where part of an application is somehow dependent on Intel and has to be re-thought and maybe some log-jams where one developer depends on another developer making an ARM version of their library or plug in - and inevitably a few casualties - but Apple have pulled this off... let's see... 6502 -> 68k**, 68k->PPC, "Classic" MacOS -> OS X (forget CPU switches - that was the biggie!), PPC -> x86-32, x86-32 -> x86-64... five times, and although that doesn't guarantee that they can pull off #6 it is still a pretty good plausibility argument.

The Windows/x86 world is hamstrung by a huge, conservative corporate sector that expects their bespoke binaries from 1995 to run in 2020 - even the one big switch (from DOS/Windows to Windows NT) took years and NT had to bend over backwards to provide a compatibility layer, which still hasn't gone away. If it were just down to a few "Top Dogs" selling "public" applications then Windows on ARM, Alpha, Sparc... maybe even Itanium (although that had other problems) might already be a thing.

The Mac has "cleaned house" every 10 years or so - the legacy support expectation just isn't there to the same degree as windows.

(** I justify including this by comparison with Windows/x86, that has an almost unbroken legacy stretching back to the days of CP/M, a contemporary of the Apple ][)

Where ARM can indeed save some complexity compared to x86: decode logic unit (ARM instructions are fixed length and have more straightforward format) and no need to support legacy crap. x86 CPUs have to support all the obsolete stuff like 32bit ISA, x87 fp operations and so on. But it’s not like ARM. plus need to do less work here.

All that extra complexity takes up space on the die, uses power (=heat, and which rises rapidly with clock speed) which are huge factors in CPU design (even desktop/workstation chips have to worry about heat dissipation). It's not about some race between the ARM and x86 instruction sets running in some theoretical scenario where everything else is equal (whatever that means) - in the real world, as long as ARM can be driven faster for the same power & thermal output and can fit more cores or more GPU shaders, hardware codecs and other acceleration technologies in the same space as x86 then ARM will have the speed advantage.

The ARM wasn't designed as "just a mobile chip" - it was designed to kick Intel's backside around the room and the first chip in 1987 did just that (...the fact that it reputedly worked first time even though they'd forgot to connect the main power supply was just a bonus) but, back then, not running x86 binaries was a deal breaker outside of the embedded/mobile market - so it was forced into a niche. What's happening today is that software is predominantly written in high-level, CPU-independent code that "just" needs recompiling, "scripting" languages that ship as source code, or sometimes bytecode that runs on 'virtual' CPUs (java, Android, even MS's current C#/.net stuff) - while windows has (finally) got wiped out in the mobile sector and is facing serious competition from Linux and web technologies. So, non-Intel processors are "thinkable" once more and various groups are making desktop/server/workstation/suprecomputer-class ARM chips once more.

 

[ruslan120]

“Dozens” might be a slight exaggeration. Yes, ARM uses separate instructions fir memory loss/store where Intel can combine this into the main instruction. But that’s about it. Both ARM and Intel will decode instructions into multiple micro-ops that are the “real” commands that CPUs understand.

Where ARM can indeed save some complexity compared to x86: decode logic unit (ARM instructions are fixed length and have more straightforward format) and no need to support legacy crap. x86 CPUs have to support all the obsolete stuff like 32bit ISA, x87 fp operations and so on. But it’s not like ARM. plus need to do less work here.

Touché.

Do you happen to know if the ISA is the public ARM ISA + proprietary instructions? Or is the whole ISA exposed?

 

[leman]

Do you happen to know if the ISA is the public ARM ISA + proprietary instructions? Or is the whole ISA exposed?

I've read that Apple has a number of custom instructions in their implementation (for example for their machine learning accelerators) which are not exposed to the developer. I would also speculate that the Mac Apple Silicon will include some additional other proprietary extensions in order to make Rosetta more efficient.

 

[Joelist]

Apple's ISA is a superset of the default ARM one. It includes all the normal ARM instructions and a lot more.

 

[Danny82]

Well said everyone.. after reading all these, I guess most are right that only Apple is able to put it off and soon others may try to follow, but it will take them awhile to follow as not only do they need to source for arm base chip, they are also reliance on windows to change.. this change really seems to me like the early challenge android was facing, the fragmented ecosystem wheres Apple full control over hardware and software is much easier to pull this off.. i really hope Apple will win this war S I'm on board this change.. ?

 

[alxhxxa]

Thank you everyone, I learned a lot !!

What I actually trying to say is that, the ARM-Macs are only suitable for enthusiast, 2 - 3 years from now, at the very least. Big guns like "Adobe Creative Cloud" are not 100% function normally within "Intel", how on earth can they flip a switch and made everything right for the ARM in 2 days ?

In Windows world, lots of companies & professionals are still rely on older version of Windows, time-tested reliability is king. I believe the same truth apply to professional Mac users as well, why would those people want to spend big on a developing system ? Computers are money making tools, not toys after all.

I'm not trying to offend anyone, but I truly believe that by the the time the ARM-Mac Eco-system is mature, if it were to happen, the first couple gens of ARM-Macs would already become some very expensive souvenirs. The spec of the ARM-Mac by then would made those 1st, 2nd, 3rd or even 4th gen ARM base Macs look pathetic

 

[Tagbert]

Thank you everyone, I learned a lot !!

What I actually trying to say is that, the ARM-Macs are only suitable for enthusiast, 2 - 3 years from now, at the very least. Big guns like "Adobe Creative Cloud" are not 100% function normally within "Intel", how on earth can they flip a switch and made everything right for the ARM in 2 days ?

In Windows world, lots of companies & professionals are still rely on older version of Windows, time-tested reliability is king. I believe the same truth apply to professional Mac users as well, why would those people want to spend big on a developing system ? Computers are money making tools, not toys after all.

I'm not trying to offend anyone, but I truly believe that by the the time the ARM-Mac Eco-system is mature, if it were to happen, the first couple gens of ARM-Macs would already become some very expensive souvenirs. The spec of the ARM-Mac by then would made those 1st, 2nd, 3rd or even 4th gen ARM base Macs look pathetic

In fact, many Intel Mac apps will just “flip a switch” to recompile for Arm. Did you see the keynote address and demo where they were showing Adobe Photoshop running on Arm? That was not Rosetta. They said that just one Adobe dev had worked on getting it ready for Arm. They also showed several Microsoft Office apps already running on Apple Silicon builds. That’s not so say that there are not some apps that will take more work but it looks like most normal apps will be relatively easy to port to Apple Silicon.

 

[sotov]

Thank you everyone, I learned a lot !!

What I actually trying to say is that, the ARM-Macs are only suitable for enthusiast, 2 - 3 years from now, at the very least. Big guns like "Adobe Creative Cloud" are not 100% function normally within "Intel", how on earth can they flip a switch and made everything right for the ARM in 2 days ?

In Windows world, lots of companies & professionals are still rely on older version of Windows, time-tested reliability is king. I believe the same truth apply to professional Mac users as well, why would those people want to spend big on a developing system ? Computers are money making tools, not toys after all.

I'm not trying to offend anyone, but I truly believe that by the the time the ARM-Mac Eco-system is mature, if it were to happen, the first couple gens of ARM-Macs would already become some very expensive souvenirs. The spec of the ARM-Mac by then would made those 1st, 2nd, 3rd or even 4th gen ARM base Macs look pathetic

You make so many false assumptions it appears your goal is to be defeatist. Don't feed the troll people.

 

[Joelist]

The fact that they already have Microsoft Office running natively on Apple Silicon is a BIG hurdle cleared. Adobe Creative Cloud is also a hurdle cleared. I am sure Apple knows what the big applications they need to have running natively out of the gate are and are working to ensure just that.

 

[Ursadorable]

The Wintel world has so much legacy baggage in terms of hardware and software, there's no way they could migrate to a different CPU architecture en mass. The only threat to Intel is AMD.. and it's a huge threat.
[automerge]1594416582[/automerge]

but Apple have pulled this off... let's see... 6502 -> 68k**, 68k->PPC, "Classic" MacOS -> OS X (forget CPU switches - that was the biggie!), PPC -> x86-32, x86-32 -> x86-64... five times, and although that doesn't guarantee that they can pull off #6 it is still a pretty good plausibility argument.

If I'm not mistaken, there was never a 6502 Mac. I believe it started with the 68000, leaving the 6502 to the Apple I, II/II+/IIe/IIc and IIgs. But yes, the Mac has been through many processor iterations, and it will easily survive the transition to ARM.

 

[peanutbridges]

In Windows world, lots of companies & professionals are still rely on older version of Windows, time-tested reliability is king. I believe the same truth apply to professional Mac users as well, why would those people want to spend big on a developing system ? Computers are money making tools, not toys after all.

When Apple started to depreciate OpenGL, that put many vendors in a bind. They long ago dropped Nvidia and consequently the whole library of support for CUDA and future Tensor acceleration. Drop OpenGL and it's just another hurdle to jump into however Metal translates it. Granted Apple is working with Blender on a Metal backend so if that works and Blender has a working stable prototype, the benefits should help others ease uncertainty moving forward. The closed nature of Metal and ditching other standards isn't buying much support from developers. Professional programs run better on Quadros or Firepros to a lesser extent via OpenGL.

Q100421: Apple deprecating OpenGL and OpenCL in macOS 10.14 (Mojave)

SUMMARY This article explains how the OpenGL and OpenCL deprecation announcement by Apple will affect the Foundry products. MORE INFORMATION Apple have announced that they are deprecating OpenGL an...

support.foundry.com

From the Blender developers in late 2019,

"If we spend most of our time ticking all the GPU support boxes, and afterwards fixing all kinds of platform specific issues, then I don’t think we are actually making the most of the limited resources we have. And for that reason I still don’t consider Metal support one of the top priorities for Cycles for funding. Because if the core rendering algorithms are not competitive then it does not matter on how many GPUs they run."

Autodesk contemplated bringing 3ds Max to MacOS when it was x86 but decided the timeline was too long. It was 2-3 years just to port over then however long to catch up. Imagine if they did that just before this ARM transition so in that case, the effort totals 3-4 years of porting only the base. That's a lot of time and resources are finite. What is the long-term return from the Mac market? Can the number of new paying users justify that cost? Everyone is asking themselves this question.

In the end, Apple will always largely be a niche. Doesn't honestly matter what they do as everyone else will tick along. There's a lot of unfounded claims of Apple leapfrogging two very experienced CPU makers and also two very experienced GPU makers. But they don't need to beat any of them as long as their own vertical platform has enough apps to lure/keep people and enough performance be it brute force or intelligent optimization/dedicated hardware to keep pace.

 

[alxhxxa]

In fact, many Intel Mac apps will just “flip a switch” to recompile for Arm. Did you see the keynote address and demo where they were showing Adobe Photoshop running on Arm? That was not Rosetta. They said that just one Adobe dev had worked on getting it ready for Arm. They also showed several Microsoft Office apps already running on Apple Silicon builds. That’s not so say that there are not some apps that will take more work but it looks like most normal apps will be relatively easy to port to Apple Silicon.

Time will tell, I sincerely hope so, even though I can't believe so
[automerge]1594426853[/automerge]

You make so many false assumptions it appears your goal is to be defeatist. Don't feed the troll people.

Apple has already won this uphill battle, before it even start... to me, that's "false assumptions" ?


I honestly believe that only 2 types of people could benefited from the early gen ARM-Macs,

1. Heavy Apple Pro app users, FCP/Logic
2. Pathological iapp users


Don't get me wrong, I have total confidence on the Apple Engineering teams, I know even the 1st gen ARM-Macs will be terrific, people talk about what kind of GPU, stuff like that, just laughable, If Apple don't have something great up their sleeves, they wouldn't dare to ditch intel; however, I firmly believe that it takes at least 3-4 years for everything done, if everything went Apple's way, if !

Ever since I brought my iphone 11 Pro, I never want to touch my old iphone 6 again, the gap is just too big ?

 

[jinnyman]

Apple market is still desirable enough for software development and publishing. If Apple's chip is on par with hypes shown in this forum, I believe many developers will continue to support Mac platform.

 

[bluecoast]

Because:

1. The Mac is a significant business in the PC world.
2. Many creative pros will only use a Mac and nothing but a Mac
3. Big software vendors have likely been told of the power and speed that’s coming down the line
4. It’ll be easier than ever for iPad first vendors to port their software to the Mac. Admittedly this will take a few years to play out.
5. And finally, no doubt they have been told that this move will result in cheaper Macs (thus an increase in sales).

 

[leman]

I honestly believe that only 2 types of people could benefited from the early gen ARM-Macs,

1. Heavy Apple Pro app users, FCP/Logic
2. Pathological iapp users

Add to this: home and office users who value better battery and power efficiency (office etc will be available on Apple Silicon from day 1), web developers, certain class of academic users... the user set will be bigger than you think

 

[Danny82]

I read everyone discussing the difficulty in the transition comparing back to the days of powerpc to intel and 32bit to 64bit.. i do agree that on the technical front it is not an easy task and it will take many years.. but do remember guys, times are different now compared to the old days.. Apple has more money for R&D and reputation to influence other tech companies.. the shift back then could have been a joke to everyone or even calling Apple late to the game because intel has won the war in CPU front.. but today, Apple influence is really strong and most tech companies will be afraid to fall behind.. even though their reliance to intel and Microsoft, and even if they wish to change they could not.. my guess is after the 2 years transition we will see a major shift in the industry.. thought I have to say.. legacy game may be dead for Apple silicon Mac and we can only hope for new games to be implemented in such a way it is easy to port for both x86_64 and Arm base..