Arm Processors with Mac Pro Level Performance Possible Today

[robbieduncan]

I think virtual desktops are the future
Thin clients. Citrix equals safe computing in the workplace and no virus's or computers down
storage and datacenter all move into the cloud

So ARM or Intel Macs will not make a difference
everything Citrix or virtual desktop. So macOS is in trouble in the corporate world of computing.

I work for a large US bank. We have used virtual desktops for about 5 years (exclusively: no one has a non-virtual desktop). We used to use Citrix, now we use VMWare Horizons. Either way I can work from any platform anywhere in the world. I have travelled on business trips taking only an iPad with a keyboard without issues.

 

[chucker23n1]

I work for a large US bank. We have used virtual desktops for about 5 years (exclusively: no one has a non-virtual desktop). We used to use Citrix, now we use VMWare Horizons. Either way I can work from any platform anywhere in the world. I have travelled on business trips taking only an iPad with a keyboard without issues.

We've been promised the thin client revolution since the 1990s' Sun network computing, and it has happened in some places, but not all.

Apple in particular isn't going to be interested in that; they want to sell a rich client-side experience to justify getting high-end devices. Otherwise, you might as well get a Chromebook or Android tablet. So if thin clients do become more common, it won't be Apple pushing for that.
[automerge]1583834046[/automerge]

Citrix equals safe computing in the workplace and no virus's or computers down

It's only "safe" if you, the user, are considered inherently not trustworthy, and in contrast, the IT department is inherently trustworthy.

Is IT more likely to be competent at keeping computers safe? Sure. But they also don't have your interests in mind any more than HR does. So it may be "safe", but also creepy.

 

[robbieduncan]

We've been promised the thin client revolution since the 1990s' Sun network computing, and it has happened in some places, but not all.

Apple in particular isn't going to be interested in that; they want to sell a rich client-side experience to justify getting high-end devices. Otherwise, you might as well get a Chromebook or Android tablet. So if thin clients do become more common, it won't be Apple pushing for that.

We are big into the BYOD world. I see so many people with iPads. The ability to use Remote Desktop on them well is just another factor in people buying them for themselves for office use

 

[MikeZTM]

Yeah. Losing virtualisation would be a massive loss for me, longer battery life doesn't mean much (I rarely use my laptop without a charger), and quieter fans while nice aren't a must-have.

I guess Apple could launch ARM Macs alongside Intel, starting at the lower end. But it'd be hard for me to see it as anything other than a step backwards.

ARM support virtualization out of the box. Just grab a Pi and see how you can run KVM on it.
[automerge]1583851505[/automerge]

Honestly, I’d rather have Intel compatibility at this point. I don’t really need more performance out of my system. In fact, I’m happy with pretty much everything my iMac does. I don’t sit down and think “man, I wish my processor was 2x as powerful.” I sit down and enjoy my speedy iMac and I’m happy.

The only reason I’d upgrade is to be able to update to the latest macOS, as I get obsoleted.

You do not care about 2x performance but there's people do.
And what about 1x performance but 2x the battery life? Or much thinner?

You basically loss nothing if the performance goes up.
You are not pushed to upgrade. You can still upgrade at your own pace, regardless of the current gen CPU performance.

 

[_Refurbished_]

I get this. It makes practical sense.

However for laptop users we should also get increased battery life.

I’d bet that this is the bread and butter of switching.

If Apple can tout 20 hour battery life with improved performance, it will be out with the old and in with the new.

 

[falainber]

We are big into the BYOD world. I see so many people with iPads. The ability to use Remote Desktop on them well is just another factor in people buying them for themselves for office use

What do call "office use"? To me it's when you sit at the desk and work. In this situation, the tablet is the worst device one can imagine (well, the phone would be worse). 34" monitor is what I prefer (or a couple of 27" monitors).

 

[robbieduncan]

What do call "office use"? To me it's when you sit at the desk and work. In this situation, the tablet is the worst device one can imagine (well, the phone would be worse). 34" monitor is what I prefer (or a couple of 27" monitors).

I see them a lot in meetings. People take them with them and can stay aware of urgent emails, see their calendars, take notes etc. So in the office but not at their desks

 

[chucker23n1]

ARM support virtualization out of the box. Just grab a Pi and see how you can run KVM on it.

I fail to see how KVM helps me virtualize Windows on x86.

Whether Windows on ARM can run on Apple Ax at all, much less without a special licensing deal with Qualcomm, is unclear.

 

[CarlJ]

Whether Windows on ARM can run on Apple Ax at all, much less without a special licensing deal with Qualcomm, is unclear.

I'm curious, why would they need licensing from Qualcomm to run Windows on ARM?

 

[Falhófnir]

I fail to see how KVM helps me virtualize Windows on x86.

Whether Windows on ARM can run on Apple Ax at all, much less without a special licensing deal with Qualcomm, is unclear.

Ultimately I expect it will depend on how valuable Apple sees Bootcamp for Windows on Arm (and whether WoA actually gains any significant traction). The primary reason for bootcamp is to run Windows software that may or may not work with WoA anyway, in which case it's of little value. But if it gains significant traction Apple might want to facilitate bootcamp, in which case it's for them to work with Microsoft to develop a version of Windows which will work with their chips. There's little reason I can think of for Microsoft to say no if Apple approach them.

 

[chucker23n1]

I'm curious, why would they need licensing from Qualcomm to run Windows on ARM?

(I'm not a low-level expert, mind you.)

TL;DR: because ARM, unlike x86, doesn't offer much in the way of a hardware abstraction layer.

So the thing about x86 is that, for historic reasons, it's become highly standardized (see "IBM-compatible PC" and "Hardware Abstraction Layer / HAL"). So whether AMD makes a CPU or Intel does, or whether Asrock designs a motherboard or Apple, the hierarchy is the same. You have, for example, a firmware that's almost invariably either BIOS (increasingly rare these days) or EFI/UEFI. Therefore, all Apple had to do to make Windows run on an x86 Mac ("Boot Camp") was to allow it in the firmware (which at the time a BIOS compatibility module, because Windows wasn't fully compatible with EFI yet), and to provide appropriate drivers for Apple-specific hardware (such as their trackpad). These days, there's even less work involved — Apple was early in using EFI, but almost everyone does now.

That's not the case for ARM at all. Yes, the instruction set is specified by Arm Holdings, but everything around it isn't. A Raspberry Pi, an iPhone, and a Microsoft Surface Pro X all use ARM, but they're otherwise quite different. All supported versions of Windows 10 on ARM use Qualcomm CPUs, and only Qualcomm CPUs. Yes, the programs they run are compiled to 64-bit ARM regardless of Qualcomm (well, either that, or they're 32-bit x86 and emulated), but drivers and firmware are Qualcomm-specific. Your non-Qualcomm ARM CPU may work, or it may not, and just because it does work doesn't mean Qualcomm is amused about that. Even if an ARM MacBook boots Windows on ARM out of the box (and it probably won't), Qualcomm's legal team is going to fight that tooth and nail, because that's not what Microsoft and Qualcomm agreed to.

Unless, of course, Apple, Microsoft, and Qualcomm want to come to an agreement.

ARM simply never had the need or stakeholders for a common hardware platform. On the contrary, each chip designer is happy to build their own proprietary and deliberately incompatible approach. Including, of course, Apple.
[automerge]1583873715[/automerge]

Ultimately I expect it will depend on how valuable Apple sees Bootcamp for Windows on Arm (and whether WoA actually gains any significant traction). The primary reason for bootcamp is to run Windows software that may or may not work with WoA anyway, in which case it's of little value.

Yup.

Compared to Microsoft's first shot at Windows on ARM, they've added emulation this time, but there still doesn't seem to be that much traction. It mostly feels like the Surface Pro X exists not so much because they believe in the product, but because they wanted to fire a warning shot at Intel to get their act together.

The developer story for Windows on ARM is pretty poor.

But if it gains significant traction Apple might want to facilitate bootcamp, in which case it's for them to work with Microsoft to develop a version of Windows which will work with their chips. There's little reason I can think of for Microsoft to say no if Apple approach them.

My expectation is they would need to bring not just Microsoft but also Qualcomm to the table. The new, emulator-enhanced Windows on ARM is entirely Qualcomm-specific. They'll want a licensing deal for their secret sauce, regardless of whether Apple finds that sauce useful.

 

[cmaier]

(I'm not a low-level expert, mind you.)

TL;DR: because ARM, unlike x86, doesn't offer much in the way of a hardware abstraction layer.

So the thing about x86 is that, for historic reasons, it's become highly standardized (see "IBM-compatible PC" and "Hardware Abstraction Layer / HAL"). So whether AMD makes a CPU or Intel does, or whether Asrock designs a motherboard or Apple, the hierarchy is the same. You have, for example, a firmware that's almost invariably either BIOS (increasingly rare these days) or EFI/UEFI. Therefore, all Apple had to do to make Windows run on an x86 Mac ("Boot Camp") was to allow it in the firmware (which at the time a BIOS compatibility module, because Windows wasn't fully compatible with EFI yet), and to provide appropriate drivers for Apple-specific hardware (such as their trackpad). These days, there's even less work involved — Apple was early in using EFI, but almost everyone does now.

That's not the case for ARM at all. Yes, the instruction set is specified by Arm Holdings, but everything around it isn't. A Raspberry Pi, an iPhone, and a Microsoft Surface Pro X all use ARM, but they're otherwise quite different. All supported versions of Windows 10 on ARM use Qualcomm CPUs, and only Qualcomm CPUs. Yes, the programs they run are compiled to 64-bit ARM regardless of Qualcomm (well, either that, or they're 32-bit x86 and emulated), but drivers and firmware are Qualcomm-specific. Your non-Qualcomm ARM CPU may work, or it may not, and just because it does work doesn't mean Qualcomm is amused about that. Even if an ARM MacBook boots Windows on ARM out of the box (and it probably won't), Qualcomm's legal team is going to fight that tooth and nail, because that's not what Microsoft and Qualcomm agreed to.

Unless, of course, Apple, Microsoft, and Qualcomm want to come to an agreement.

ARM simply never had the need or stakeholders for a common hardware platform. On the contrary, each chip designer is happy to build their own proprietary and deliberately incompatible approach. Including, of course, Apple.

This doesn't make a lot of sense to me. The parts of the qualcomm chipset that are unique (that are not defined by the ARM ISA) would be addressed by drivers. Apple could simply provide different drivers, just like it does for bootcamp. I must be missing something

 

[MikeZTM]

I fail to see how KVM helps me virtualize Windows on x86.

Whether Windows on ARM can run on Apple Ax at all, much less without a special licensing deal with Qualcomm, is unclear.

Windows on ARM already runs on Pi (Pi 4 via KVM).

And Intel also need BSP. It just we only have Intel and AMD for x86 Windows so people almost forgot about them. Only recently AMD need special scheduling for high performance brought this back to people's sight.

 

[DeepIn2U]

I work for a large US bank. We have used virtual desktops for about 5 years (exclusively: no one has a non-virtual desktop). We used to use Citrix, now we use VMWare Horizons. Either way I can work from any platform anywhere in the world. I have travelled on business trips taking only an iPad with a keyboard without issues.

I had worked for the worlds largest agriculture company and it too would use VMWare Horizons .... allowing those still working for the company after a regional office closing to work from anywhere in the world ... even off their iPads. I was quite impressed.

Banks by nature should ALWAYS have a virtual environment, not just from a security standpoint yet also helping in situations where a pandemic fall-back plan to keep the business running is feasible.

 

[AlexEmme]

Oh another transition 😍😍😍
(Sorry, i'm a retro-guy and I'll continue to use my protools system on Mac Os9 🧡 )

 

[robbieduncan]

I had worked for the worlds largest agriculture company and it too would use VMWare Horizons .... allowing those still working for the company after a regional office closing to work from anywhere in the world ... even off their iPads. I was quite impressed.

Banks by nature should ALWAYS have a virtual environment, not just from a security standpoint yet also helping in situations where a pandemic fall-back plan to keep the business running is feasible.

Indeed. We tested our remote abilities last weekend. In Europe we had almost 60,000 unique employees logged in from home on Sunday to test capacity: it was fine so we are good for our contingency planning

 

[laz232]

If Apple uses x86-to-ARM transcoding (recompiling x86 binary code to ARM and run it natively), you probably won't notice much difference. There might be some initial latency and that's about it.

As I mentioned in my post - not if the code is running in a Windows VM (E.g. EDA / CAD design software (price tag 7000 euro / without VAT, Windows only) - then recompilation is not applicable and there will no doubt be a substantial performance hit. Due to the work I do I already travel to clients with a Macbook and a Win10 small laptop - don't want to add another computer or increase my Win10 laptop size, as I really dislike using Windows10

 

[MikeZTM]

As I mentioned in my post - not if the code is running in a Windows VM (E.g. EDA / CAD design software (price tag 7000 euro / without VAT, Windows only) - then recompilation is not applicable and there will no doubt be a substantial performance hit. Due to the work I do I already travel to clients with a Macbook and a Win10 small laptop - don't want to add another computer or increase my Win10 laptop size, as I really dislike using Windows10

Back in the days all those professional software support all CPU archs including PowerPC, DEC Alpha and SPARC.
It just these two decades x86 becomes popular and they start to stick to it.
And they figured out they can charge you twice for different platform. That's how they earns money and nothing you can do about it. They are punishing people using older version of their software by locking you down to old systems (hardware and software) and push you to pay them again.

If you already carrying two computers then there's no different for you, just run windows app on windows computer and Mac apps on Mac. You do not need to increase the size of Windows computer as you can just remote to it from your Mac.

 

[smeagol]

I trust Gassée for his technical knowledge but it is still a tough call to make this switch. The server/pro market moves insanely slowly and it has to be a pretty compelling argument to make and architecture switch. It means rebuilds and re-optimizations of high end pro software which is not cheap. To make this happen is a decade+ long game which is hard to justify. I WANT it and it'd make every cross platform dev's life easier but that may not be enough.

If anyone can do it it's Apple. They successfully transitioned away from Power processors to Intel. Yes, there were a few bumps in the road but it was not a disaster.

 

[cmaier]

If anyone can do it it's Apple. They successfully transitioned away from Power processors to Intel. Yes, there were a few bumps in the road but it was not a disaster.

And from 68k to PowerPC. (Everyone always forgets that Mac has had four different ISAs - 68k, PowerPC, x86, and x86-64.)

 

[laz232]

If anyone can do it it's Apple. They successfully transitioned away from Power processors to Intel. Yes, there were a few bumps in the road but it was not a disaster.

It would require a lot of effort - and I doubt that Apple cares - their actions have shown that they give 0 shoots about the professionals (apart from a very narrow range of video producers). They prefer to make a "iPad Pro Plus" with a keyboard and not pay Intel the 300 USD per CPU (which adds about 700-800 USD to the price of a Macbook if you look at the retail to bill of materials ratio)
[automerge]1584023769[/automerge]

And from 68k to PowerPC. (Everyone always forgets that Mac has had four different ISAs - 68k, PowerPC, x86, and x86-64.)

Except that PowerPC was >>> 68k (and even then the emulator had bad performance until the PowerPC 604, where the entire emulator could fit in the cache). Similarly there was a big performance increase from PowerPC -> Intel x86, thus that took care of the emulation slow-down.
6-core i5 MacbookPro to (custom) ARM? I don't see that that won't have a large performance hit...

 

[chucker23n1]

Except that PowerPC was >>> 68k (and even then the emulator had bad performance until the PowerPC 604, where the entire emulator could fit in the cache). Similarly there was a big performance increase from PowerPC -> Intel x86, thus that took care of the emulation slow-down.

I don't remember the bad performance so much as I remember the dreaded error 11. For years, there were edge cases where the 68k emulator failed, and would crash things.

6-core i5 MacbookPro to (custom) ARM? I don't see that that won't have a large performance hit...

I don't think so. It'll improve. Six-core mobile Intel chips at 45W only reach a single-core score of about 1100, yet a six-core iPhone at ~5W beats 1300. That's about a tenth the TDP, yet 20% better single-core performance. If you take a 7W TDP Intel CPU, it's way, way, worse at 758. So right now, compared to the current shipping MacBook Air? 75% better at single core.

Sure, Ice Lake would do better. Tiger Lake, even better. But "a large performance hit"? I don't see that at all.

To your point, I do not think that ARM Macs would be a significant almost two-fold boost compared to x86 the way Intel vs. PowerPC or PowerPC vs. 68k were. But leaving aside emulation, I don't think performance would be worse. Probably slightly better.

 

[AppliedMicro]

One of the biggest hurdles moving to ARM is going to be those that will want to run Windows and other OSes alongside macOS through virtualisation. Or native Windows via Boot Camp.

I would be unhappy for this very reason - but how big a fraction of their user base is it?

I don't see any major change to architecture for a few more years - the Mac Pro has just been released. It must have cost Apple a fortune to design and support it, and to announce the move to ARM would royally hack off those people that have spent the money on the likes of a Mac Pro or even the iMac Pro

The previous Mac Pro had a shelf life of 6+ years.

What if this iteration of the Mac Pro is just that: A stop-gap solution on their way to transitioning to ARM? The most powerful Mac they've ever made - but at the same time the pinnacle for x86-64 Macs - which is never going to be surpassed (on this architecture) in processing power? Intended to be powerful - and well-timed - to "tide over" these professionals for a couple of years, until their 3rd party software is optimized well enough for a new architecture?

 

[chucker23n1]

I would be unhappy for this very reason - but how big a fraction of their user base is it?

I'm guessing 10%, if even that.

(I'm one of those.)

Now if you take the pro Macs, though, that percentage certainly goes up. The MacBook Pro continues to be a very popular cross-platform development laptop. I wouldn't be surprised if 20% of MacBook Pro owners have a Windows partition or Windows VM.

I don't know what I'll do if this happens. Will there be a Boot Camp with Windows on ARM? Probably not, because ARM is less standardized than x86, so the effort would be higher (and there might be licensing requirements Apple may not want to go through), and the benefit would be lower than it was in 2006, when getting as many people as possible to the Mac was a much higher priority. Will Windows run fast enough in emulation? For casual apps, maybe. Almost certainly not for development. Will enough of my development have shifted to run natively on the Mac by then? Probably not either.

The previous Mac Pro had a shelf life of 6+ years.

What if this iteration of the Mac Pro is just that: A stop-gap solution on their way to transitioning to ARM? The most powerful Mac they've ever made - but at the same time the pinnacle for x86-64 Macs - which is never going to be surpassed (on this architecture) in processing power? Intended to be powerful - and well-timed - to "tide over" these professionals for a couple of years, until their 3rd party software is optimized well enough for a new architecture?

Yeah, that's possible.

If ARM Macs are a real thing, I don't expect an ARM Mac Pro soon; instead, I think it makes more sense to start with the MacBook Air, and then slowly work your way up, maybe over the course of two to three years.

 

[CarlJ]

What if this iteration of the Mac Pro is just that: A stop-gap solution on their way to transitioning to ARM? The most powerful Mac they've ever made - but at the same time the pinnacle for x86-64 Macs - which is never going to be surpassed (on this architecture) in processing power? Intended to be powerful - and well-timed - to "tide over" these professionals for a couple of years, until their 3rd party software is optimized well enough for a new architecture?

There's another way to look at it - by the time they switched to Intel, the PowerPC chips were less optimal in nearly every way, and the transition was fairly rapid. Currently, the x86-64 chips are really good at a lot of things (like high performance apps and running Intel-based VMs), so it doesn't need to be as rapid of a transition. Anything coming out of Apple or the Mac App Store goes through their Bitcode process, so it's compiled/customized for the right/current CPU, and they got fat binaries (containing code for multiple architectures) worked out pretty well last time. So it's much less of a hard either/or transition. They could come out with one new Mac using an ARM chip, like a revived MacBook, that would hugely benefit from the increased power savings, and then slowly work their way up the line over numerous years. By the time they got to the Mac Pro, Intel compatibility might be less vital. In that case, the current Mac Pro wouldn't be a stop-gap per se, it'd live a full life, and would be followed by an ARM replacement, likely using the same chassis and much of the same technology.