ARM based MacBook Pro discussion [merged]

[jsnuff1]

Same here, so I really hope they'll do one final update to the Intel 16" this fall.

Unless they are finally giving us a 10nm CPU then you might as well go with an ARM...intel CPUs are now essentially on 5 year old tech, while everyone else is already on 5nm, and is the reason why Apple is going ARM in the first place.
[automerge]1591891842[/automerge]

Maintaining a running Macbook is costing me $100+ a month, so I switched most of my 24/7 stuff to a $70 Raspberry PI running ARM - it's fanless, it's silent, it's cool and it's cool, I love it

100 bucks a month on maintenance...what are you smoking bro? My car does not cost that much for maintenance. I think the most I have spent on my 2017 MBP in the last 3 years is 100 bucks worth of air dusters and alcohol wipes lol

 

[DanMan619]

However, I think this transition will be very good for the Mac, and not just in processor performance. Apple is still mostly a phone company. I think the T2 chip is largely a kludge to bring over some of their phone innovations. macARM will mean we'll see a lot more innovation and development in traditional computers again. Personally, I think its a very exciting time!

This is what has me excited as well. If Apple goes all in on ARM it drastically increases the incentive and ability for them to really start innovating on Mac again, which they haven't really been doing in recent years. It will hopefully put more control back in their hands to swing for the fences with these new ARM computers.

 

[09872738]

I agree, in fact game developers may be what drives this transition since the iOS market is so big. They either have an iOS version that they can port up to the Mac (...)

Sorry to bust your bubble, but you fall victim to a fallacy like so many: Please note that the CPU architecture has close to nothing to do with ease of porting. iOS apps, including games, usually contain close to no CPU/platform specific code. So porting is not a question of CPU architecture, but of software, i.e. frameworks, APIs and toolchain available.

In other words: macARM will NOT make it easier to port iOS apps to the Mac just because both are running on ARM.
A suitable toolchain, maybe including a cross compilation, will. Also note that a toolchain like that does not necessarily have to run on ARM

 

[gnomeisland]

Sorry to bust your bubble, but you fall victim to a fallacy like so many: Please note that the CPU architecture has close to nothing to do with ease of porting. iOS apps, including games, usually contain close to no CPU/platform specific code. So porting is not a question of CPU architecture, but of software, i.e. frameworks, APIs and toolchain available.

In other words: macARM will NOT make it easier to port iOS apps to the Mac just because both are running on ARM.
A suitable toolchain, maybe including a cross compilation, will. Also note that a toolchain like that does not necessarily have to run on ARM

I never said there would CPU specific code involved. The tools to port to one platform will be very similar to the other. That isn’t about CPU code and it is true of the developer tools Apple has focused on (the last few years especially).

To port their programs, in the case games, it will require a certain amount of work. That is true now since Apple uses different APIs (Metal vs DirectX prob being the biggest for games). Right now, porting games to Mac is a lot of work with little payoff. Now, porting to one mean using the tools will essentially make it run on both platforms, the more niche mac OS and the larger (but with a lot of legacy, slower systems) iOS.

In theory this has been somewhat true for a while but now it should also mean less time optimizing code (which is system/API dependent) because there will be so many more similares between the two systems.

 

[psingh01]

That reminds me. Microsoft put out the Surface Pro X last November. It had a custom ARM chip (the SQ1 chip) that Microsoft made in collaboration with Qualcomm that was based on Qualcomm Snapdragon chips (used in Android phones). It ran the full desktop version of Windows 10 coded for ARM. So that makes me wonder what that could imply for Boot Camp with Windows 10 being available on ARM Macs fairly early on (though maybe not day 1) in the transition since the ARM version of Windows 10 already exists.

Well it might be possible, but it’s the software that runs on windows that’s important. Not the OS itself. For all I know there is a better chance at native Mac versions of the software than Windows ARM versions ? software developers won’t waste much time on that platform if they can’t make money on it and Microsoft isn’t likely to force them to support it like Apple would.

 

[09872738]

To port their programs, in the case games, it will require a certain amount of work.

Potentially, yes, of course. However, when porting an iOS app it does not matter if your Mac runs on ARM or Intel. Whatsoever.

(...) or justify porting across the macARM because the jump to iOS ARM is a relatively small (...)

?is a false hope. There is close to no difference between macARM and macIntel. Same APIs. Same toolchain. Same everything. For the vast, vast majority of apps no difference in effort required at all.

Don't get me wrong, I am not at all opposed to macARM and of course would love some innovation (and quite a bit of back to the roots when it comes to software). Unfortunately the underlying CPU won't make a difference there.

 

[mscice]

I will not be surprised to see apple CPU's compete with AMD and Intel in near future... and apple will do all they can to be #1

I see no point in buying any laptops right now from Apple... wait for ARM... first release of ARM will blow everything out ... imagine how many sites will benchmark / review new ARM macbook and compare it with the rest... Apple will do all they can do have best possible results and not show a EPIC fail

 

[kaans]

If Steve Jobs was at the helm, most definitely

But with Tim Cook, best case scenario, we'll suffer the growing pains

 

[leman]

To port their programs, in the case games, it will require a certain amount of work. That is true now since Apple uses different APIs (Metal vs DirectX prob being the biggest for games). Right now, porting games to Mac is a lot of work with little payoff. Now, porting to one mean using the tools will essentially make it run on both platforms, the more niche mac OS and the larger (but with a lot of legacy, slower systems) iOS.

Porting most games from Windows to macOS is arguably easier than porting them from macOS to iOS. Different input method, different performance characteristics of the GPU... even when you stay within Metal only, the tiled renderers on OS are a very different beast that support a different subset of the API and need to be treated in a special way to get high performance.

With porting, most of the work is done at the design stage. If you design your software in a platform-agnostic way (which is not that difficult), porting can be relatively easy. But of course, to do this you need competent and motivated software designers with cross-platform experience AND competent managers...

Anyway, as @09872738 says, the instruction set architecture plays a lesser role here, unless one is specifically relying on features of a specific CPU. 64bit ARM and 64-bit macOS have the exact same rules for laying out binary data, so if you have clean code that compiles and runs on modern Intel macOS, it will most likely compile and run without modifications on ARM macOS.

 

[pshufd]

Porting most games from Windows to macOS is arguably easier than porting them from macOS to iOS. Different input method, different performance characteristics of the GPU... even when you stay within Metal only, the tiled renderers on OS are a very different beast that support a different subset of the API and need to be treated in a special way to get high performance.

With porting, most of the work is done at the design stage. If you design your software in a platform-agnostic way (which is not that difficult), porting can be relatively easy. But of course, to do this you need competent and motivated software designers with cross-platform experience AND competent managers...

Anyway, as @09872738 says, the instruction set architecture plays a lesser role here, unless one is specifically relying on features of a specific CPU. 64bit ARM and 64-bit macOS have the exact same rules for laying out binary data, so if you have clean code that compiles and runs on modern Intel macOS, it will most likely compile and run without modifications on ARM macOS.

We use ICC where I work. My understanding is that AMD uses it as well. Intel also makes some pretty high-performing vector libraries. You're not going to get those on ARM. Does ARM have 512 bit vectors now?

 

[leman]

We use ICC where I work. My understanding is that AMD uses it as well. Intel also makes some pretty high-performing vector libraries. You're not going to get those on ARM. Does ARM have 512 bit vectors now?

ARM has SVE (scalable vector extensions). SVE instructions are SIMD vector width agnostic, meaning that SVE code will automatically take advantage of whatever vector width the CPU supports. CPU makers can choose from 128 to 2048 bit vector units, so there is definitively some space to grow. No idea about hardware support though, this tech is fairly new. Current Apple ARM CPUs support 128bit vectors if I’m not mistaken (no SVE support yet). Personally, I think that vector-width-agnostic programming has tremendous potential, but it will most likely take a few years until we get there in practical terms.

On the Intel side only workstation CPUs and new 10nm mobile support AVX-512. AMD is similarly stuck at AVX2. Talking about high performance vector libraries, Apples accelerate framework was quite good last time I checked. Of course, depends on what you need to do. As far as I know, ICC produces suboptimal code on AMD, but that might have changed in the meantime.

All in all, it all very much depends to what Apple has in store (if anything). If they can deliver ARM CPUs with 512 bit simd, competitive clocks and higher IPC within a reasonable timeframe (say 3 years), it could be very impactful for pro workflows. If what they present is just a slightly faster A13 with only Neon, not as interesting (but good enough for consumer).

 

[pshufd]

ARM has SVE (scalable vector extensions). SVE instructions are SIMD vector width agnostic, meaning that SVE code will automatically take advantage of whatever vector width the CPU supports. CPU makers can choose from 128 to 2048 bit vector units, so there is definitively some space to grow. No idea about hardware support though, this tech is fairly new. Current Apple ARM CPUs support 128bit vectors if I’m not mistaken (no SVE support yet). Personally, I think that vector-width-agnostic programming has tremendous potential, but it will most likely take a few years until we get there in practical terms.

On the Intel side only workstation CPUs and new 10nm mobile support AVX-512. AMD is similarly stuck at AVX2. Talking about high performance vector libraries, Apples accelerate framework was quite good last time I checked. Of course, depends on what you need to do. As far as I know, ICC produces suboptimal code on AMD, but that might have changed in the meantime.

All in all, it all very much depends to what Apple has in store (if anything). If they can deliver ARM CPUs with 512 bit simd, competitive clocks and higher IPC within a reasonable timeframe (say 3 years), it could be very impactful for pro workflows. If what they present is just a slightly faster A13 with only Neon, not as interesting (but good enough for consumer).

It's not optimal as ICC only cares about Intel chips latencies and clocks. But I heard that AMD uses ICC for benchmarks.

 

[dagmar10]

If ARM MBPs are in the near future, is it still worth considering an Intel MBP 13" 2020? Would that be like buying a PowerPC Mac right before that transition?

I'm curious about this too. If I buy a 2020 machine will I be able to use it normally for a few years and it will keep its value for when I inevitably sell it. I read somewhere that when PowerPC Mac users that bought right before the Intel switch they had a near useless machine -- I don't want this happening if I purchase a 2020 MBP/MBAir.

 

[pshufd]

I'm curious about this too. If I buy a 2020 machine will I be able to use it normally for a few years and it will keep its value for when I inevitably sell it. I read somewhere that when PowerPC Mac users that bought right before the Intel switch they had a near useless machine -- I don't want this happening if I purchase a 2020 MBP/MBAir.

Exactly what I would do. I usually run three to four versions behind on macOS. Used equipment is fine for me but I'd probably pick up a new Mac Pro or MacBook Pro at the end of the line.

 

[spook]

Exactly what I would do. I usually run three to four versions behind on macOS. Used equipment is fine for me but I'd probably pick up a new Mac Pro or MacBook Pro at the end of the line.

I was looking at buying a 16" Macbook Pro, but with this news I am not sure what to do. We are talking over £3,000 that could be obsolete in 6 months. I'm holding off till 22nd for clairty, but I am now considering a new XPS

 

[pshufd]

I was looking at buying a 16" Macbook Pro, but with this news I am not sure what to do. We are talking over £3,000 that could be obsolete in 6 months. I'm holding off till 22nd for clairty, but I am now considering a new XPS

It won't be obsolete. You could always just throw Windows or Linux is Apple drops support. I doubt that they will do that given that they have a pro community willing to spend $10 - $20 thousand on those machines.

 

[imrazor]

It won't be obsolete. You could always just throw Windows or Linux is Apple drops support. I doubt that they will do that given that they have a pro community willing to spend $10 - $20 thousand on those machines.

Linux support on the 16" MBP is pretty dicey ATM. Support is very much a mixed bag, thanks to the T2 chip. I've had quite a bit of trouble getting Fedora 32 working properly, though I haven't tried any other distros yet.

 

[JPack]

I was looking at buying a 16" Macbook Pro, but with this news I am not sure what to do. We are talking over £3,000 that could be obsolete in 6 months. I'm holding off till 22nd for clairty, but I am now considering a new XPS

The writer of the Bloomberg article revealing the ARM transition suggested not to buy a Mac until the new ARM devices are released. The current products won't be obsolete in 6 months, more like 18-36 months, but that's still a relatively short period of time.

 

[DanMan619]

The writer of the Bloomberg article revealing the ARM transition suggested not to buy a Mac until the new ARM devices are released. The current products won't be obsolete in 6 months, more like 18-36 months, but that's still a relatively short period of time.

Yeah, the 2020 Intel Macs won't be immediately obsolete, but they won't go the 5-7 year distance a lot of people aim for when buying these machines. Comparing to the PowerPC to Intel transition. The transition was announced in 2005, Intel Macs rolled out all throughout 2006. PowerPC machines still got one more compatible Mac OS (Snow Leopard) that came out in 2007, and third party software developers still supported their softwares for Snow Leopard for however long they did (it varied but probably wasn't more than a year or two). This was also the last Mac OS with Rosetta (emulated PowerPC softwares) support if you were on a new Intel Mac but still needed access to PowerPC software. I'd imagine we'll see a fairly similar timeline for Intel to ARM, which lines up with 1.5-3 years like you said. I'm really curious to see how resale value for these last Intel Macs will be impacted.

 

[Falhófnir]

Yeah, the 2020 Intel Macs won't be immediately obsolete, but they won't go the 5-7 year distance a lot of people aim for when buying these machines. Comparing to the PowerPC to Intel transition. The transition was announced in 2005, Intel Macs rolled out all throughout 2006. PowerPC machines still got one more compatible Mac OS (Snow Leopard) that came out in 2007, and third party software developers still supported their softwares for Snow Leopard for however long they did (it varied but probably wasn't more than a year or two). This was also the last Mac OS with Rosetta (emulated PowerPC softwares) support if you were on a new Intel Mac but still needed access to PowerPC software. I'd imagine we'll see a fairly similar timeline for Intel to ARM, which lines up with 1.5-3 years like you said. I'm really curious to see how resale value for these last Intel Macs will be impacted.

I think Apple might give them 5-7 years OS support from launch, at least in some form (say Intel support goes up to 10.17/18 but they keep security patching it for several more years?) Almost certainly developer support will very quickly drain away once the transition launches, though. Apple easily have the resources to do this (unlike 2005) and it saves them some face on dropping support for relatively new machines, even if they're only technically supported.

 

[ipos]

are there any advantages of ARM chip over intel chip?

 

[pshufd]

are there any advantages of ARM chip over intel chip?

For Apple, it means ending reliance on Intel and AMD for processors.

 

[matram]

are there any advantages of ARM chip over intel chip?

Generally ARM is more power efficient. So either you could build a MBA with same power and perhaps double the battery life. Or you could in the same thermal capacity (think MBP) provide more power.

In the longer term Apple can provide more dedicated silicon integrated with the CPU as they today do with SSD controllers. This could result in very good performance for specialized use cases.

 

[mguzzi]

I haven't read through all of this thread but how will Apple's move to ARM effect the ability to run VMWare with multiple Windows and Linux Virtual machines. Also - I remember when AMD was new and how much trouble their was with software compatibility and driver issues.

 

[matram]

I haven't read through all of this thread but how will Apple's move to ARM effect the ability to run VMWare with multiple Windows and Linux Virtual machines. Also - I remember when AMD was new and how much trouble their was with software compatibility and driver issues.

It will, virtualization requires an Intel CPU for code compiled for Intel, else we are talking about emulation which has a high performance penalty.

We will probably know soon if it will be impossible to run a Windows VM or if it will just be very slow.