Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Silicon: The Complete Guide
- Thread starter MacRumors
- Start date
- Sort by reaction score
"I just wonder whether an arm based mac makes any sense in a market that is getting smaller and more niche." => It doesn't
"I think we are more likely to see iPads morphing into the low cost computing line than MacBooks getting cheaper with arm chips." => I agree
If I recall correctly, the PowerPC Macs were extremely competitive against Intel machines, when they first came out. And then IBM soon enough failed to deliver on continuing strong year-over-year upgrades to performance - PowerPC chips got years of modest speedbumps while Intel CPUs continued a steady climb that passed the PowerPCs. Apple bet heavily on IBM's promises, and IBM failed them. If they hadn't, we might still be on PowerPC hardware. Apple has a long memory for things like that. That's why they really prefer to be on hardware that they control, like the A-series chips for their phones.
(Such long memories are, I think, also part of why the only path to iOS apps is through Xcode and the App Store, a path that Apple completely controls - from what I recall, a lot of major players ended up using Metroworks C for their Mac apps, and then when Apple wanted to move the Mac platform forward, it was extraordinarily hard to get those major players to move off of Metroworks - it's why Carbon stuck around for so long.)
Not so. Check "Surface Pro X app compatibility", it says "x64 apps won't work". There was an article explaining how emulating x64 on ARM64 is a virtually unsolvable problem. Windows general manager Erin Chapple revealed for zdnet early 2018 that this issue will never be resolved. The technical hurdle is just too big to address. And that stance still hasn't changed, and will probably never change. Never say never, but extremely unlikely.
No question that Microsoft is trying to move to the ARM direction, but it's not an ideal platform for professional apps. At least not until AutoCAD, Solid Works, Premier, Illustrator, Photoshop and others are rebuilt and proven on ARM.
It's unclear whether Apple will release macOS for ARM, like a MacBook Airm. They might stick to the decision to strictly separate the iPad Pro and the macOS platforms.
They need to be careful, there's a very fine line between a successful platform and a failed one. Microsoft has kept failing repeatedly. It's a very tough problem. What matters is not just how good the product is, but the perception of the potential buyers. Professionals need to be reassured that they won't be neglected. The only way I see this happen is releasing a completely redesigned Intel MacBook Pro and a MacBook Airm at the same time, hinting that both platforms are taken seriously. They cannot release an ARM Mac and not reassure people.
Thanks, I didn't know that about x64, I was mistaken.
I agree with everything you said. I think there are many factors for why Windows on ARM has failed before, some of which you addressed. One factor that isn't discussed often is cost of software updates.
Most software vendors of professional software aren't going to release ARM-native apps of old versions. For example, maybe Dassault will release SolidWorks 2021 as a native ARM app. But they aren't going to go back and update SolidWorks 2018. Many people and companies rely on several big applications like that. So a change to ARM, in effect means you have buy updates to all the software in your workflow at the same time as getting a new workstation/laptop. Setting aside the compatibility and interoperability issues and other nits that will inherently come up on such a big workspace undertaking, it also turns a relatively routine new computer purchase into a many tens of thousands of dollars expense per user.
In almost every way, the path of least resistance is to just stick with x86/x64. It's cheaper in the short term, much lower risk of incompatibility, easier roll-out, can continue using legacy software, etc.
If this happens, my take is that Apple would continue to use Intel chips on the iMac Pro and Mac Pro indefinitely. Developers would need to compile their apps for both Intel and ARM. However, this is not 2005 where that was a major burden. 32-bit support is over, Carbon is gone, and all devs are working in one IDE (XCode). Apple even fully controls the compiler at this point. The "click the checkbox" to recompile without any other work may actually be possible now.
Last edited:
Based on what? This is because you IMAGINE that the higher tiers will work better on Intel cores, but you have no evidence for this belief.
Everything we know about how Apple has performed in chip design so far suggests that they are perfectly capable of creating a many-core SoC (8, 16, 32+ large cores). Replicating cores is easy. They have a performant NoC. If they want to group cores (likely) either for L2 sharing or as chiplets, they control the OS and so can make it NUCA aware. And lower power cores means that when they're all running they'll likely run at substantially higher throughput than Intel at the same thermals.
The ONLY two things that are missing right now are
- an AVX512 equivalent. Easily solved by adding SVE.
- x86 support. Which I consider grossly overrated in importance, but if that's something you care about, you buy the new Mac Pro when it comes out, then you're set for 5 to 7 years during which whoever it is that you think won't support ARM ASAP can get their act together.
Meanwhile what you'll get from Apple controlling the SoC is a vastly faster rate of improvement. PCIe4 on the first release, with PCIe5 following soon after. USB4 very soon. No more multi-year waits to support the next LP-DDR spec. Innovations like providing HBM support (at least a pool of a few GB) for the CPU. etc etc.
That's why adobe went all in on subscription based license. They don't want to support multiple version of old software on different platform anymore and give user a benefit of always using latest version at same price they paid for one version.
And BTW PhotoShop already runs on iPad so it's proven working. Not all features are there now but the foundation is here and features are much less a issue compared to the engine of the software.
ISTR the key problem was that IBM was more interested in big metal and there was never a low-power G5 for the PowerBook.
However - it wasn't just a 'steady climb' by Intel - Intel went down a huge, expensive dead end with its Pentium 4/"Netburst" CISC architecture which was much better at heating up the room than computing. People actually started using the Pentium M mobile chip (based on the older Pentium Pro RISC-with-a-CISC-instruction-translator architecture) as the go-to chip for general purpose desktops. Intel did a massive U turn, ditched Netburst and went back to Pentium pro/M to create the "Core" series, which were much better - especially at power efficiency and muti-core processing - than what Intel had done before. The MacBooks debuted at the same time as the Core processors (and the change to Core was quite likely a major factor in Apple's decision).
The issue was that Mac OS X was a completely and utterly different operating system - essentially an updated NeXTStep and based on Unix - to "Classic" Mac OS, and Classic applications needed to be substantially re-written to run on it. "Carbon" was an API designed to partially mimic the "Classic" API to make it easier to port existing code. "Cocoa" was the new-world, NeXTStep-derived API (half the class names in the OS X API still start with "NS" to this day) and using it involved a far more substantial re-write (...including, preferably, switching languages from C++ to the not-as-similar-as-it-sounds Objective C).
I know Metroworks was the most popular C compiler/IDE for "Classic" and switching IDEs can be a royal pain - but not as much as switching to a totally alien API while also learning a language that looks like random hieroglyphics to someone familiar with C/C++. (Objective C was a fine language, but unlike C++ took inspiration for its object-oriented extensions from Smalltalk, a language designed to be used by children: in other words, You Have Been Warned
)
I'm using AWS ARM servers. They are pretty good at virtualizing ARM platform. So virtualizing Windows ARM64 is possible and there's an Apple virtualization hypervisor shipped with macOS right now.
And You can get a copy of Windows on Arm just from Microsoft Online store--purchase Windows 10 pro retail version and get the key, that's it! ARM64 version of windows is just like x86/x64 version. You are not buying x86 or arm64 version by itself. You are getting arm64 and x86 when you purchased x64 version.
Windows on arm is a all-in bet from Microsoft this time. It's not a special version. Just remember NT3.0 had PowerPC and Alpha version and x86 was added after that.
[automerge]1573068522[/automerge]
At the end of day, all libraries are built with only Apple provided system APIs anyway. You just have to check the box for all libraries and that's it. As the pointer and basic variables are exactly same size this switch is easier than switching from x86 to x64.
Last edited:
I imagine we are years away from ARM macs
IMO there are too many advantages to x86 like bootcamp and ease of cross-development.
if it did happen however there'd basically be a few year life cycle to x86 macs as i dont think apple would do a clean break unless there was maybe a (rosetta 2.0/classic environment 3.0)
I imagine if a fully ARM future is in the 10-20 year plan Apple would be more likely to do ARM co-processors that way pro's could offload video, audio or other encodings etc. which would be a unique value proposition the dispossessed pro's that are now hackintoshing won't have.
after a several years of that they will make the ARM more powerful and the x86 less so by comparison until they can fully switch.
IMO there are too many advantages to x86 like bootcamp and ease of cross-development.
if it did happen however there'd basically be a few year life cycle to x86 macs as i dont think apple would do a clean break unless there was maybe a (rosetta 2.0/classic environment 3.0)
I imagine if a fully ARM future is in the 10-20 year plan Apple would be more likely to do ARM co-processors that way pro's could offload video, audio or other encodings etc. which would be a unique value proposition the dispossessed pro's that are now hackintoshing won't have.
after a several years of that they will make the ARM more powerful and the x86 less so by comparison until they can fully switch.
Lightroom CC has a ARM version on iPad that shares same bug with their desktop counterparts. I guess they will release that to Windows also pretty soon.
[automerge]1573069327[/automerge]
By SPEC2006 A13 peak single core is slightly faster than a 3900X single core and only slightly slower than a 9900k at only 5W! Those x86 cores have to run at 30W-40W to get this single core performance.
I guess this is good enough and 8 large cores with fan to make peak performance last forever at 40W will be well faster than any MacBook we have today. The i9 MacBook Pro is only running at 2.3GHz and way slower than a 9900k.
Last edited:
Isn’t a custom GPU running on Metal the first obvious step? They’ve been transitioning developers over to Metal for years now, and it’s the GPUs in most Macs which are the weakness. Intel has been allowed to stand still in part because CPUs aren’t that slow for the average consumer, and haven’t been for some time.
Here is a classic example of Schrodinger's Game, which simultaneously exists on a platform and yet somehow isn't a "real" game at the same time, even if such games send hundreds of millions of dollars in revenue to developers.
PC-first games are always going to be tailored to the PC. If you enjoy playing PC games, buy a PC or your console of choice. The economics aren't there to bring them to the Mac.
However, according to Statista.com, at current count, there are a staggering 903,489 apps classified as 'games' on the iOS app store.
(Sure, most of them are probably derivative and terrible, but there are a surprising number of console-quality titles that span multiple genres, and some that created genres on their own like Pokemon Go)
I think right now, it’s just for their own device, isn’t it? Have any large vendors announced support for Windows on ARM systems that they plan to ship? And, if they can carve out a nice profitable niche for themselves (because these systems woouldn’t come with the Intel tax), they’d keep making them.
They’re likely to have issues with EVERY die shrink now because Intel’s processors are more complex than Qualcomm’s or Apple’s, it comes from a requirement to be everything to everyone. For example, in every Intel chip, there’s a considerable amount of space specifically for handling instructions of variable lengths.
[automerge]1573070816[/automerge]
Apple lost the professional market in general a LONG time ago. The only professionals that still use macOS SHOULD be the ones that are using FCPX and Logic Pro. I say SHOULD because those working on 5 year old machines waiting for Apple to provide them with a solution they want to buy at a price they’d want to buy it at SHOULD be on some other platform being more productive, but... what can ya say?
How fast do these few occasionally used apps need to run? Will they run on a 10 year old PC? If so, they probably will likely run well enough for occasional use under a slow x86 emulator on a fast ARM processor. Or from a cloud virtualized desktop.
Only power users of pro windows apps need a fast x86 macBook, and those apps will run tons faster from remote desktop VMs in the cloud (AWS EC2 z1d's instances, et.al.) than on any lightweight laptop, Dell or Apple. MacBook Airs aren't meant to compete in the AlienWare market. But outsell them by magnitudes.
" The only professionals that still use macOS SHOULD be the ones that are using FCPX and Logic Pro"
And Xcode. iOS developers still constitute the largest base of professional users for Apple's ecosystem.
I don't think we're going to see and Arm-based Mac anytime soon for several reasons.
Is Apple going to maintain two different MacOS versions, begging devs to develop bug free apps for both architectures, for an Arm MacBook with a ridiculous marketshare? I don't think so
IMHO we'll never see an Arm MacOS device. iPad and (Arm) iPadOS will continue to grow in performances, features and versatility to meet the needs of the vast majority of users. Smart Keyboard, mouse and external memory support are a big step in this direction, and future developments will end up confining Mac and (Intel) MacOS to pro users
- Many posts on MacRumors forums show that even 64 bit transition, despite devs had plenty of time to recompile their apps, is going to be a pain for many users. After two years of reminders and announcements from Apple, a lot of apps are still 32 bits only, even if this transition should be much less demanding than a complete change of architecture
- Surface Pro X is the perfect example of x86 on Arm emulation limits. Reviews showed that emulated apps are painfully slow and consume a lot of power. Even though I think Apple can do a better job than Microsoft in this respect, there will always be a significant overload due to emulation
- Maybe some apps can be recompiled with a simple click on the checkbox, but MacOS itself needs to be rewritten from scratch... Kernel extensions, device drivers etc will need much more work than a simple click, requiring a lot of effort (and time) from both Apple and manufacturers
- Every MacOs 10.x.0 version has always been buggy and rough, despite most part of code is exactly the same as the previous version. It could take light-years to have a stabile and reliable Arm MacOs...
- Mac Pro is gong to be x86, so it wouldn't be a fast transition as it was in PowerPC era. Carry on two different architectures would be time and resource demanding for both devs and Apple, leading to an exponential growth of bugs and incompatibilities. And nowadays Mac represents less than 10% of Apple revenue
Is Apple going to maintain two different MacOS versions, begging devs to develop bug free apps for both architectures, for an Arm MacBook with a ridiculous marketshare? I don't think so
IMHO we'll never see an Arm MacOS device. iPad and (Arm) iPadOS will continue to grow in performances, features and versatility to meet the needs of the vast majority of users. Smart Keyboard, mouse and external memory support are a big step in this direction, and future developments will end up confining Mac and (Intel) MacOS to pro users
Changing from x86 32bit to x64 is more demanding than changing from x64 to arm64 because of different pointer sizes and many other size issues.
Emulation is temporary for old abandoned app to work. Most x64 app can compile to arm64 within a week if they have full source code access.
macOS kernel already runs on arm--that is called iOS and since compiler is the same clang LLVM any system library written in C++/ObjC shouldn't be a problem. And you do not need to rewrite OS to support a new arch, Linux runs on arm with minimal changes mostly not for Cpu arch but for special manufacture's add-ons. Modern OS are built mostly using high level language instead of assembly and compiler will do the magic.
A13 is already reaching 9900k performance why not replace Mac Pro with ARM chip in the future?
But actually there's a catch: Why we need a Mac? Make Xcode and FCP runs on iPad and that's it!
If my iPad is already faster than a MacBook Air why the hell do they need to improve the Mac? Just axe it and make it history.
Last edited:
There are tasks for which a GPU is better than a CPU. But instead of replacing the CPU on a computer with a GPU, I’d rather have both, so tasks can be assigned to whichever processor handles them best. There are tasks for which ARM is better than x86. But instead of replacing x86 with ARM, I’d rather have both, so tasks can be assigned to whichever processor handles them best. Perhaps T2 chips on Macs are a step in this direction.
and if the intel switch never happened so people didnt experience the benefits of x86 i'd be giving those same reasons why going only ARM would make sense.
im just saying now you have almost a decade and a half of a large portion of your userbase utilizing benefits of being x86 whether your talking bootcamp, crossover, virtualization like vmware/parallels, not to mention wine/cider and other ports like QT.
many people even non-techie have benefited by the intel switch and don't know it.
that will be a hard drug to get off without a transition period.
thus why i envision co-processors for 3-5+ years before a full switch.
its not about raw speed, mac market share hasn't increased enough to warrant many developers to make the new development challenges worth it and on top of that even compared to last time with rosetta, Apple doesn't appear to care as much as they used to about the mac to do a new "rosetta" to help with the switch.
plus this time, transitive technologies cant save their bacon cause they became debunked after IBM bought them and ive seen no other promising ARM to X86 instruction set emulators that appear to be efficient enough at the moment.
Indeed they may come up with a Macbook Air ARM version to test the waters, and for day to day general use of browser-based apps, it will probably work brilliantly. There are also plenty of "good enough" apps for typing word documents, spreadsheets, editing of short videos and photos yet it all abruptly ends if you get involved with some real heavy workloads. If you take a look at the MacPro line it is quite obvious that the ARM chips are an add on part and not really a full replacement of the Intel... at least not yet.
If I understand correctly, apps submitted to the App Store these days are uploaded with bitcode, a sort of semi-compiled non-architecture-specific copy of the code, which Apple can finish compiling for various platforms, complete with architecture-specific optimizations. They use this to be able to download only an x86 (32-bit) version or x86_64 (64-bit) version of the application to a customer, as needed, rather than having to have "fat binaries" on everyone's system (which contained all possible versions of the executable code). (I think they also did this for 32-bit vs. 64-bit iOS apps.)
There's nothing that says they couldn't use this same system in the future to automatically compile/deliver both Intel and ARM-specific versions of an app, from the same resources and bitcode, without the developer having to go through any special steps. Presumably they could release a small tool (if they haven't already) to process a compiled application down to a specific architecture version, for all the developers who don't go through the Mac App Store.
And since it's an intermediate layer in the Xcode stack, I'm guessing it ought to work for any language Xcode can compile (Swift, ObjC, C, C++), which would cover the majority of Mac developers. Assuming any 3rd-party frameworks you use are also compiled the same way.
Compiling for the same app on two different operating systems is pretty hard to get right. Compiling for the same operating system on different architectures is considerably easier (I do it on Linux all the time).
If Apple can engineer better, faster, more efficient CPUs.. why?