What are the benefits of Apple switching to Arm Macs?

[Krevnik]

There are tons of threads in this sub forum discussing exactly this...

To put it short: Apple is not switching to ARM. They are switching to in house designed chips since they have better tech than Intel.

I find these sort of semantic debates funny. It's like saying AMD wasn't x86 (Intel's ISA), or Intel isn't x64 (AMD's ISA). ARM is the architecture, Apple Silicon is the implementation of the architecture. Apple doesn't want to call it ARM precisely because they've been adding custom ASICs and compute units to their SoCs, stuff that competitors aren't investing in. The emphasis Apple wants is on what they bring to the table as a whole, where ARM is just a piece of the puzzle.

It's still an ARM SoC though.

The x86_64 chipmakers are doing power and energy cores in future too.

And I wonder how far out that is. Why wouldn't Apple jump now if they can be years ahead of AMD or Intel in this space? They've already been delivering this sort of power scaling with ultralight laptop level performance for a couple years.

Unfortunately, Googling for news on this front mostly brings up ARM developments, not x86 ones.

Part of this however is misunderstanding what TDP means. Intel defines TDP as power under sustained load running on full cores at base frequency. To put it differently: they guarantee that if you load up all the cores with work and limit maximal power dissipation to the TDP, the CPU will run at least at base frequency. This is why TDP is more of a marketing term rather then a technical one. Turbo boost is completely circumstantial.

But what TDP means is irrelevant to the point being made. Part of the problem with Intel at the moment is they are effectively trying to push boost clocks higher to remain competitive. So if you want to see the better performance, you have to rely on the skyrocketing power consumption to do so. Nonsense like that is partly why we have a battery in the 16" MBP that's at the limit allowed for air travel, with a power supply just below the 100W limit that USB-C PD supports. Even then, the CPU alone can spike into the 90-100W range under some loads, essentially demanding the whole output of the power brick to itself. Yeesh.

AMD has recently started releasing some incredibly competitive chips as of late. Some are just plain beastly, such as the Threadripper, and the SOC for the Xbox Series X.

The problem that AMD hasn't really addressed yet though is Machine Learning. Sure, you can just throw that on the GPU, but there's better ways to do it.

Yeah, on-die ASICs are definitely a strength of Apple's SoCs at the moment.

But I'd argue that while AMD pulls down less power at the high end, it's not doing as well as Intel during low loads/idle with the 3000-series. My 3600 pulls about 8W on idle. While the i5 in the Mac Mini pulls a little under half that. I'll admit that's comparing Windows where nothing is running vs macOS. Still, that's not exactly a great place to be if trying to court Apple. Nor does shipping chips that require a microcode update to not boost to full clockspeed and draw a lot of power every time Windows sneezes. That was a fun couple of months...

The thing is, AMD's chips are very competitive. But I'm not entirely sold on the reliability of AMD's design validation process right now.

 

[HowardEv]

I noticed in that chart comparing New Apple Silicon Macs to Laptops and Desktops, there is nothing at the Desktop level of power. The new chips will apparently all be laptop TDP 45W, meaning the new Macs won't be "desktops" anymore, they probably will all have internal battery and use USB-C power adapters rather than plug into the wall. I predict a two piece form factor, a display and a processor, with new wireless connection.

 

[JMacHack]

I just throw this here: with apple silicon, macOS will eventually reach iOS state, completely locked down to Mac App Store (or a universal App Store). Users will unable to install anything outside App Store. All power user features will be gone or completely hidden behind Xcode or advanced apps. I just cannot believe Apple will maintain the semi-openness of current macOS in the near future despite apple’s words.

Which would be tantamount to killing the goose laying the golden eggs.

Likewise, nothing with Intel's processors prevents this.

MacOS' open ecosystem is what facilitates iOS' closed ecosystem. The whole reason XCode hasn't been released for iOS is because it would be embarrassingly easy to sideload apps and jailbreak that way.

 

[farewelwilliams]

Probably dreaming on that.

Apple in 2011 Considered Collecting a 40% Fee From Some Subscription Apps

Apple once considered taking a 40 percent cut from some subscription apps, according to documents shared today by the House Judiciary Committee (via...

www.macrumors.com

Any CPU package bill of materials (BOM) cost reduction Apple will likely shuffle either into some other component of higher BOM or just pocket it. So perhaps larger SSD capacity (at same Apple $400/TB rate) , micro-LED screens , etc.

At the higher end of the Mac product line up the volume of Macs is low enough that there my not be a drop in BOM costs for the CPU packages. A-series are cheaper in part because Apple selling 10's of millions per year of them. 27" iMac , iMac Pro , Mac Pro are probably 2 , if not 3 , orders of magnitude lower in volume. That isn't going to reduce BOM costs at all (fixed costs up and volume cost spread over much smaller number of units ).


Same thing on the iPhones when Apple drops Qualcomm later. Probably going to cost just as much.

Nope.

https://twitter.com/i/web/status/1286898274427756544

 

[leman]

Don't necessarily need physically unified memory to do that. Especially in the narrow corner case of video decode.

Apple's Afterburner card sits on a discrete PCI-e add-in card and does four 4K HDR decode streams. The PCI-e bus in a Mac Pro is not "narrow" in any sensible characterization. Perhaps the MBP 16" and its dGPU.

A GPU that can deal with a virtual memory mapping can just just map in the video data and pull it in through the hardware decode. Doesn't necessarily have to be copied to VRAM at all except to the video out framebuffer post decoding.

That more a matter of what Apple hasn't worked on as opposed can't be done.

16 lanes of PCI-e 3.0 is just under 16GB/s. That is indeed very narrow in a world where regular system DRAM offers bandwidth of 60GB/s or more, and GPU VRAM even 300+ GB/s. This is not a problem if the data can be pre-loaded into the GPU memory (like is the case with games) or if the work done by the GPU takes much longer than it’s needed to copy the data itself. But if you are working with multiple high-res video sources that are way to large to be preloaded, it might take a couple of ms until enough data has been copied to the GPU to continue smooth playback - enough delay to cause noticeable stuttering.

Besides, imagine the benefits of being able to work on the same data on the CPU and the GPU simultaneously. I am sure this will open up completely new applications.

 

[Shirasaki]

Which would be tantamount to killing the goose laying the golden eggs.

Likewise, nothing with Intel's processors prevents this.

MacOS' open ecosystem is what facilitates iOS' closed ecosystem. The whole reason XCode hasn't been released for iOS is because it would be embarrassingly easy to sideload apps and jailbreak that way.

Not really.

Intel Mac not having universal App Store is because, well, intel, and x86 architecture. Developer needs to compile stuff for both Apple Silicon and x86. With Apple silicon in mac, developers don’t need that anymore, and Apple can do the most conversion hardwork themselves, saving devs time and money into porting their iPadOS and iOS apps to Mac.

Apple not releasing Xcode in iOS could be because coding and testing requires iOS to have some sort of developer mode, similar to the one on Android. They haven’t found a compromise yet to make developing apps on iOS feasible. Once found, Apple will push Xcode to iOS and devs can do their projects on the go.

While MacBook is high profit margin, people are more than willing to spend $2000 or even $3000 for a premium iPhone instead of a MacBook with the same price tag, because it is a lot of computing power with you all the time. MacBook with active cooling etc can push better performance, but nobody gonna carry MacBook around to browse Facebook Twitter etc. but on the phone it is fine. iPad, while still being extremely hampered by iOS, looks like the computing device Apple wants everyone to use. Portable, much easier to carry around than MacBook, aesthetically pleasing for a lot of people, and versatile to an extent. Why buy a MacBook when iPad can do everything you need, without fan noise?

 

[MevetS]

I just throw this here: with apple silicon, macOS will eventually reach iOS state, completely locked down to Mac App Store (or a universal App Store). Users will unable to install anything outside App Store. All power user features will be gone or completely hidden behind Xcode or advanced apps. I just cannot believe Apple will maintain the semi-openness of current macOS in the near future despite apple’s words.

Why does this have anything to do with ASi? If they wanted to lock it down that could be done with any processor.

Never gonna happen.

 

[leman]

I find these sort of semantic debates funny. It's like saying AMD wasn't x86 (Intel's ISA), or Intel isn't x64 (AMD's ISA). ARM is the architecture, Apple Silicon is the implementation of the architecture. Apple doesn't want to call it ARM precisely because they've been adding custom ASICs and compute units to their SoCs, stuff that competitors aren't investing in. The emphasis Apple wants is on what they bring to the table as a whole, where ARM is just a piece of the puzzle.

It's still an ARM SoC though.

I find them funny too But you know, this is all about the connotation. When we talk about “x86” average computer user familiar with the term thinks “Windows”, “performance”, “pc”, “games”, “productivity”. When we talk about ARM, the same user things “phones”, “slow”, “low power”, “longer battery”, “gadgets”. Not to mention that ARM can also mean “cores designed by ARM”. You can see this effect clearly in these forums, where people constantly make these kind of associations.

I think that by trying to distance themselves from the ARM moniker, Apple is first and foremost trying to distance themselves from these connotations. Over time, as more high-performance ARM CPUs will become available, these connotations will change, but for now, Apple‘s marketing department doesn’t want people to think that they are using phone CPUs to run their desktops.

Case in point: we are talking about Intel CPUs and AMD CPUs. We don’t say ”my gaming PC has an x86 CPU” - because this statement does not contain any useful information. ”ARM CPU” is similar, only worse, because it can literally mean anything while connotating “my PC is slow and crappy”.

 

[JMacHack]

people are more than willing to spend $2000 or even $3000 for a premium iPhone instead of a MacBook with the same price tag

... or because phone plans will bill you monthly instead of a lump sum like a computer.

 

[leman]

And I wonder how far out that is. Why wouldn't Apple jump now if they can be years ahead of AMD or Intel in this space? They've already been delivering this sort of power scaling with ultralight laptop level performance for a couple years.

Unfortunately, Googling for news on this front mostly brings up ARM developments, not x86 ones.

Try “Intel Lakefield”. It’s already out. And it’s really bad. Packaging technology is interesting though.

To be honest, I don’t see much hope for this approach in the Windows world. Efficient big.little utilization requires good management by the OS. Apple have spent years introducing various multitasking and power assertion APIs. They have tons of data on how to identify performance critical and non performance critical threads. A Mac app written in the last couple of years using Apple frameworks provides enough information to the OS scheduler. On Windows, they will just ask people to mess around with thread affinity APIs - good luck with that - or OS will make some random choices. It will get better over time of course, but I bought it will ever work as well as on ARM platforms who have their roots in ultra-mobile computing and hence paid much more attention to power usage annotations.

 

[MaxinMusicCity]

example of what i believe is speculation already in this thread

1. The switch is attributable to Intel delays (and if so that it was an sufficient or important consideration)

2. That Apple will release macs more frequently

Meaning they may be obsoleted more frequently forcing one to have to buy new more frequently to keep up with macOS updates. Ultimately pricing many OUT of the already expensive Mac market. On the other hand, if having their own chips could allow them to lower the Mac prices...

 

[Joelist]

Apple laid out the reasons for the switch already - they want to do things on the Mac platform that only can be achieved using their own SOCs.

Oh, and ARM is NOT an architecture. All it is is an instruction set. The architecture of the A Series is 100% Apple in origin and if it resembles anything it actually somewhat resembles the Intel Core 2 microarchitecture.

 

[thejadedmonkey]

Try “Intel Lakefield”. It’s already out. And it’s really bad. Packaging technology is interesting though.

To be honest, I don’t see much hope for this approach in the Windows world. Efficient big.little utilization requires good management by the OS. Apple have spent years introducing various multitasking and power assertion APIs. They have tons of data on how to identify performance critical and non performance critical threads. A Mac app written in the last couple of years using Apple frameworks provides enough information to the OS scheduler. On Windows, they will just ask people to mess around with thread affinity APIs - good luck with that - or OS will make some random choices. It will get better over time of course, but I bought it will ever work as well as on ARM platforms who have their roots in ultra-mobile computing and hence paid much more attention to power usage annotations.

The Windows scheduler isn't half as bad as you make it out to be. Sure it's crap on Intel PCs, but I'd wager that's more a function of Intel than Windows. Intel uses time to drop back to a low power state to differentiate between their Gold and U3 CPUs, at least. And I also know that on mobile, Windows Phone had really good battery life, even comparing favorably to Apple at the time.

Edit: I should add, Windows can eek out 15+ hours on my Surface Go if I'm using it as an ebook reader, vs 7 hours of "light" use, or 5 hours of video conferencing from a web browser. So really, it's not bad.

 

[Realityck]

Apple laid out the reasons for the switch already - they want to do things on the Mac platform that only can be achieved using their own SOCs.

Today, we’re going to tell you about some really big changes. How we’re going to take the Mac to a whole new level. From the very beginning, the Mac redefined the entire computer industry. The Mac has always been about innovation and boldly pushing things forward, embracing big changes to stay at the forefront of personal computing. The Mac has had three major transitions in its history. The move to PowerPC, the transition to Mac OS X, and the move to Intel. And now it’s time for a huge leap forward for the Mac. Because today is the day we’re announcing that the Mac is transitioning to our own Apple Silicon. When we make bold changes, it’s for one simple yet powerful reason, so we can make much better products. When we look ahead, we envision some amazing new products, and transitioning to our own custom Silicon is what will enable us to bring them to life.

If you tie in your comment to Tim Cook's future viewing statement marked in bold, it still leaves us all very much in the dark because it doesn't guarantee that the transition is to the same product line up as now or that Apple won't produce something new.

 

[leman]

The Windows scheduler isn't half as bad as you make it out to be. Sure it's crap on Intel PCs, but I'd wager that's more a function of Intel than Windows. Intel uses time to drop back to a low power state to differentiate between their Gold and U3 CPUs, at least. And I also know that on mobile, Windows Phone had really good battery life, even comparing favorably to Apple at the time.

It’s definitely not bad. Windows thread scheduler and memory management is state of the art. But I was talking about asymmetric CPU setups, with big and little cores. You can’t treat them like your regular symmetric CPU Setup and Windows is simply not prepared for that.

 

[jvlfilms]

Optimization? It is pretty likely that 7,1 Mac Pro that is being lamented on has an Afterburner card in it. It is this specific corner case hardware decoding that is the difference there. The FPGA in the Afterburner card is more than capable of doing the work if it was programmed to do this specific case. Apple didn't do the work so it doesn't. That isn't an optimization, that is just lack of putting in the effort.

Apple wants folks to transcode into ProRES RAW .

Very similar issues around Apple putting in work (an/dor explicitly contracting for the work to be done) around the hardware en/decode of the AMD ( and Nvidia ) GPUs that is present.


Apple picked 4K 4:4:2 probably in part to record higher gamut color video without incurring the cost of storage that 4:4:4 would. H.265 (HEVC) version 2 has seven profiles 4:2:0 (1) 4:2:2 (2) , 4:4:4 ( 4 ) . Most of the hardware implementations aimed at high end video skipped the middle 2. For the "princess and the pea" colorists 4:4:4 is more likely a pressing issue and for economic "good enough" the 4:2:0 works better.

Apple picking what Canon picked isn't so much of an optimization and as the two happen to pick the same narrow range "landing spot".

Great if happen to pick the Cannon R5 and have an iPad Pro . Not so great (and far closer to average ) if happen to pick from a much wider array of cameras out there. if recording where Apple is herding folks too ( ProRes RAW) it isn't much a gap at all.

I don't really understand the comparison you're making.

The video is showing the ability to playback and edit native 4K files from this particular camera inside various machines including a MacBook Pro 16" and 7,1 Mac Pro in a variety of software - Final Cut Pro, DaVinci Resolve and Adobe Premiere.

The use of an Afterburner card is irrelevant in this case as these video files aren't ProRes. We're talking about the CPU and GPU capabilities of $3,000+ machines with dedicated graphics cards and Intel CPU's. The footage is stuttery on even the $15,000 Mac Pro

Yet, the iPad Pro can handle these files fine - on a $30 app called LumaFusion (must have app). The reason is clearly due to the GPU and CPU capabilities of Apple Silicon because there's simply no reason why an AMD GPU with 8GB of dedicated memory can't playback these files at 1/2 or 1/4 resolution.

 

[Krevnik]

I think that by trying to distance themselves from the ARM moniker, Apple is first and foremost trying to distance themselves from these connotations. Over time, as more high-performance ARM CPUs will become available, these connotations will change, but for now, Apple‘s marketing department doesn’t want people to think that they are using phone CPUs to run their desktops.

I agree with this, and partly the point I was trying to make.

Case in point: we are talking about Intel CPUs and AMD CPUs. We don’t say ”my gaming PC has an x86 CPU” - because this statement does not contain any useful information. ”ARM CPU” is similar, only worse, because it can literally mean anything while connotating “my PC is slow and crappy”.

It's more you said "Apple is not switching to ARM". Yes, we can be more specific, and nuance is useful here because of how Apple's designs diverge from Qualcomm/Samsung designs, but this just reads poorly because it isn't more specific, nor nuanced. It reads more like someone declaring a giraffe isn't a mammal.

Try “Intel Lakefield”. It’s already out. And it’s really bad. Packaging technology is interesting though.

Hmm, uses Atom for the low power cores, and only has a single large core. Yeah, no wonder it's not very good. So it seems like Apple is still years ahead here.

Hell, I'm running across references to issues folks are hitting with Samsung's big.Little chips, where it turns out they used ARMv8.0 on the little cores but v8.1 on the big cores, so you can crash if your thread gets moved to one of them and you used an ARMv8.1 instruction. Whoops?

Still, I'd love to see what Apple can do with their low power cores in higher end hardware. A Mac Pro that doesn't use more power than an incandescent lightbulb while doing nothing would be a nice win, IMO.

 

[Shirasaki]

Why does this have anything to do with ASi? If they wanted to lock it down that could be done with any processor.

Never gonna happen.

Then just check out all of those issues with MacBook Pro having T2 chip. That chip is Apple designed silicon controlling camera, fingerprint reader, Secure Enclave etc.

 

[Joelist]

Then just check out all of those issues with MacBook Pro having T2 chip. That chip is Apple designed silicon controlling camera, fingerprint reader, Secure Enclave etc.

Good point. T2 is indeed Apple Silicon so current Macs already are hybrid machines. And T2 is the foundation for the hardware security of the Mac. It controls Secure Boot in a way which makes a LOT of hacks physically impossible (such as remotely loading an altered OS). It stores security information like fingerprints in a separate memory space only it can access and it alone handles authentication. If you are using FileVault and Data Encryption it performs the encryption with its own AES engine that has its own individual keys, so that remote hacks can't grab anything intelligible.

Once they go completely to Apple Silicon the T2 is still present as a fully isolated block in the SOC. I know there were people who got upset at T2 because of it making repairs/upgrades by third parties hard but the tradeoff is Macs have truly powerful security.

 

[dmccloud]

I noticed in that chart comparing New Apple Silicon Macs to Laptops and Desktops, there is nothing at the Desktop level of power. The new chips will apparently all be laptop TDP 45W, meaning the new Macs won't be "desktops" anymore, they probably will all have internal battery and use USB-C power adapters rather than plug into the wall. I predict a two piece form factor, a display and a processor, with new wireless connection.

Nothing has been announced yet in terms of specific processors for the new Macs yet, so there isn't anything to even use as a point of comparison. We do know that Apple will have a different series of processors (leaving the A-series for iPhone and iPad devices) for the new Macs, but there's nothing specific enough to even guess at where the new processors will slot into the laptop/desktop matrix.

 

[Joelist]

Nothing has been announced yet in terms of specific processors for the new Macs yet, so there isn't anything to even use as a point of comparison. We do know that Apple will have a different series of processors (leaving the A-series for iPhone and iPad devices) for the new Macs, but there's nothing specific enough to even guess at where the new processors will slot into the laptop/desktop matrix.

I expect that one reason we will be seeing a new SOC family for Macs is the much higher thermal envelope they can play in. Another will be that they probably will have some coprocessor blocks that are Mac exclusive.

 

[Wolfpup]

Why did Apple switch to Arm Macs?

Still new to Arm

What are the benefits of Arm Mac Computers?

(Rather learn research on here, Macrumors)

The only benefit to US (contrary to wacky claims about it) is Apple can more easily run iOS programs on macOS. They could do that anyway, and who knows if they will regardless, but they haven’t, and this would make it much easier for them since the hardware will be identical.

the benefit to Apple is their chips will be 1/10th the price, but they won’t pass that savings on to us, as they never do. Of course developing them is incredibly expensive, but they’re already doing that For their iOS devices, and they’re now able to just reuse the same stuff

I’m not a fan of this, although I would like to be able to run iOS programs in macOS....and macOS needed touchscreen support years ago.

I’d be vastly happier if they went the other way, keeping the macintoshes on x86, and switching the iPad over to x86 also, and letting it run some macOS stuff.
[automerge]1596171813[/automerge]

Larger margins for Apple

yeah, that’s the real reason. We won’t see the savings.

 

[smoking monkey]

The only benefit to US (contrary to wacky claims about it) is Apple can more easily run iOS programs on macOS. They could do that anyway, and who knows if they will regardless, but they haven’t, and this would make it much easier for them since the hardware will be identical.

the benefit to Apple is their chips will be 1/10th the price, but they won’t pass that savings on to us, as they never do. Of course developing them is incredibly expensive, but they’re already doing that For their iOS devices, and they’re now able to just reuse the same stuff

I’m not a fan of this, although I would like to be able to run iOS programs in macOS....and macOS needed touchscreen support years ago.

I’d be vastly happier if they went the other way, keeping the macintoshes on x86, and switching the iPad over to x86 also, and letting it run some macOS stuff.
[automerge]1596171813[/automerge]

Huh? The only benefit?
It's been stated and it's common consensus at this point that the Arm Chips will run more efficiently and far cooler meaning better potential battery life, less or no fan noise and better thermals meaning less thermal stress to the insides of the computer.
That's a massive deal if you're using a lap top computer.

 

[dmccloud]

This still doesn’t explain why Apple wouldn’t use their own good processors regardless of intel.

Look at Apple's track record since Steve Jobs returned to the company. They usually are not the first to enter a new market, but they look at what works and doesn't work, then place their own spin on things. It happened with the iPod, iMac, iPhone and iPad, but the common thread is they waited until they had a device that could mark a major shift in that market segment. Now you have machines from Samsung, Lenovo, and even Microsoft that are running Windows on an ARM-based processor (usually Qualcomm parts), none of which are really making much noise in the PC market, and are actually known more for their hit and miss software compatibility. I have a feeling that the iPad (specifically the iPad Pro) has been a testing ground for ideas they want to incorporate into new versions of Mac OS on Apple processors, and they waited until they had the basics in place to make the formal announcement.

 

[satcomer]

All I know is the Intel screwed the peach and now has lost the hard core gamers around the Gamer World, especially this year and jumped to AMD processors!