Many games still aren't available on Mac, and things like Wine are not nearly as clean of an experience as just installing games in native Windows. I'd guess switching to ARM will set back the progress made in Mac gaming as well.
People mostly dual-boot so they can access software they need that isn't on macOS, but it isn't available on ARM Windows either so why would Apple care to foster a nascent competing platform?
Additionally, the importance of this use-case (from Apple's perspective) has decreased significantly since they originally developed Boot Camp. While I'm sure that many people do run Windows on Mac hardware, the percentage of Mac owners who do this is likely dropping significantly (if I remember correctly, there was a statistic about this in a comment on another thread about ARM Macs).
Boot Camp does not work with Linux on Intel Macs, so the question in this case would be whether the low-level bootloader would allow unsigned/third-party code or whether it would be locked-down like on iOS devices.
SIP is a simple toggle (for what it's worth, this is the same thing on Windows to be able to run arbitrary UWP applications). Jailbreaking is very much not, as it requires an actual exploit.
Apple has target margin levels, they may increase slightly but most of the savings will go to customers. Apple understands that most customers can’t justify the higher cost of Macs despite the additional value, they will for sure attempt to use ARM CPUs to grow Mac market share.
Given that my iPhone's A11 Bionic beats my desktop's i5-7400 on Geekbench's single-core benchmark, I would expect that a newer desktop-optimized Apple chip would destroy most Intel laptop chips.
Often when Apple shifts the rest of the industry follows within 5 years (touch screen smart phones, ultrabooks, high DPI displays). Although Apple has done architecture shifts before with no impact to the rest of the industry (68k -> PPC -> Intel), I wonder if this will be different.
I read in another thread that volume is what propelled x86s popularity. Intel’s difficulty with 10nm has created an opening for other fabs and other architectures to catch up. If ARM can be shown to work as a desktop processor, why wouldn’t HP, Dell, Asus and the rest make ARM laptops?
The Windows PC industry has historically been, and still is, much more dependent on 3rd party software than the Apple ecosystem. So performance will not be good without porting 3rd party software to ARM.
It's similar problem to how Chromebooks have still not really displaced Windows laptops except at the very low end of the market.
Although the significance of this seems lower now than at any time I can remember as more happens in a browser than ever. Windows 10 apparently is fully ARM compatible with a built-in x86 emulation layer, so perhaps it could work.
>why wouldn’t HP, Dell, Asus and the rest make ARM laptops?
They already do, but they aren't as popular or as capable for two reasons:
The first is that only Apple can take advantage of any speed or power management improvements that a switch to a different CPU might provide, and that's because software development for Windows is very much not slaved to the App Store in terms of application development. That wasn't through lack of trying by Microsoft in the early 2010s (Metro/UWP and the disaster that was Windows 8/8.1), but as a consequence of that failure Windows applications aren't developed with a mobile-first paradigm or cross-platform support, so their ability to be managed by the OS is more limited.
The second (which is also an implication of the first) is that the only time ARM can actually outperform x86 designs is when a single-digit-watt power envelope is required. In Apple's case, it's required because they refuse to implement a proper thermal management solution in their laptops, and as such using processors that can reach 4 GHz for about 5 seconds until they hit 100C and have to aggressively throttle back are a waste (and their fans spin up to 100%, and excessively heat up the keyboard and the base of the laptop).
By going to a processor type that will ultimately be significantly slower but much more consistent in terms of performance, it's possible to both use alternate cooling solutions (like fully passive cooling) and provide a much more consistent user experience. And provided that the GPU is any better than Iris IGP, and that video encoding is otherwise accelerated (i.e. Quick Sync), professional (non-developer) users will still notice an upgrade.
And sure, while developers will lose in terms of raw performance, Apple knows that developers that go out of their way to use their products are more than willing to make compromises and concessions to use their ecosystem. On the other hand, those who are concerned about performance don't (and never) buy Macbooks.
I hope Windows 10 ARM will be able to to run on these. I think that might be the push W10 ARM needs to get recognized and have more native applications. It would be somewhat ironic too!
Performance per watt is the biggest advantage, Apple will get fast laptops with extremely long battery life. Cost per performance is a very close second, ARM CPUs are a fraction of the cost of Intel CPUs.
It’s also speculated that with active cooling ARM will enable high performance Macs at a lower price point. It’s very unclear how well they will be able to match AMD and Intel at the highest performance levels.
Perhaps a combined SoC with both ARM and x86-64 cores, with independent switching like Apple already has for GPU integrated/discrete graphics switching. I think this would be a convenient middle way through this compatibility challenge.
Can anyone explain why Apple would be dropping Intel and doing ARM instead, instead of just, you know, doing their own thing? I find it hard to believe either that ARM is the (globally optimal) pinnacle of instruction set / CPU architecture possible, and that a $1tn company can afford to spin up its own CPU production line (competitor to Intel, a $250b company) but not its own CPU design department (competitor to ARM, bough by SoftBank for $30b) ?!?
An ISA doesn't just exist in a vacuum. There is also the compiler toolchain, runtime libs, developer mindshare and years of very smart people thinking about the ISA.
A $1e12 company doesn't have to pick the "pinnacle" of ISA/CPU possible. They just have to pick one that is good enough to meet their financial goals.
I think you fail to grasp the sheer amount of time to create a new instruction set, easily a decade to get to the level ARM is now. Not just the hardware, but debuggers, tooling etc etc. Can't see that paying off
Also, the kernel codebase, which is shared with iPhones, will be able to target a single architecture
The instruction set matters less for performance than people think. But it does matter a lot for the ecosystem: low-level hyper-optimized libraries for certain operations (think BLAS), tooling, developers understanding how to best write for it, etc. Creating all of that will take a long time no matter how much money you throw at it. For ARM, that ecosystem is probably the second best in the world, and in particular Apple has been building it up for themselves for more than a decade since the first iPhone.
Additionally, sharing the ISA means that the ARM reference cores will always be available as a backstop if Apple's custom chips start running into trouble for whatever reason. IIRC, Qualcomm was doing a lot of hopping between full custom cores and ARM reference ones depending on which one happened to work better some particular year.
ISAs are somewhat orthogonal to performance these days, so this move is more or less switching to Apple CPU designs. Apple has rapidly iterated and improved in-house designs since basically the A4 iPad.
Apple effectively has its own proprietary ARM designs, best in the world, and production relationships with chip manufacturers to make them in huge volume. Dumping that for some clean sheet of paper design without real world customers, tool chains, and manufacturing to try to gain a tiny percent in performance and/or cost would be immensely foolish.
It’s been speculated Apple could even license the x86 instruction set to include a x86 coprocessor in every one of their designs.
Because Apple's business is not doing "their own thing".
Apple's business is taking something that already is proven to work, and integrate it into something that makes money.
ARM is proven to work. It was designed for phones, tablets, routers, everything.
That means you could go to a factory like TSMC and they can reuse in excess of 95% of their hardware chip templates. You can reuse work done in iOS hardware too.
We are talking about freaking less than 10 nanometer scale designs!! The transistor's layout at those scales is not even square but calculated with quantum effects with very complex software, and retouched by hand with very expensive feedback using electron and X-rays microscopes.
That also means you can reuse the compilers that already exist, the transpilers and the toolchain in general.
Hardware designs need literally years of testing in very sophisticated machines(supercomputers) before you use them. Remember that if you make a mistake your entire die is wrong, million of dollars go to the toilet. And you always make mistakes before the design is mature.
It is not software that you could update easily and cheaply.
If you were to do your own thing , your chips would cost you 100x to 1000x to make just in money. It would also cost you orders of magnitude in time to market.
Of course a $trillion company can waste money away(until there is no more money left) but probably they want their money to give profit and create wealth instead of consuming it.
Docker on ARM Linux doesnt use x86 emulation, I'm not sure why Docker on ARM macOS would.
edit: I am aware not all Docker containers run on aarch64. But that has been changing already, and presumably would accelerate if Apple goes all in on ARM.
I suppose now that Edge is cross platform there's less need to spin up a windows VM to test in IE/Old Edge.
I do worry about software that is highly tuned with hand-rolled assembly (looking at you various video/photo/audio processing apps) and if this is going to be another transition that is mostly seamless, but with enough edge cases to leave people slightly bloodied.
I would think that most video, and photo processing apps are using the GPU rather than custom assembly on the CPU since you have hundreds of cores on a GPU.
>Starting in 2021, Kuo says that all new Mac models will be equipped with Apple processors, and that it will take 12 to 18 months for Apple to transition to an all-Arm lineup.
So there's not going to be a Xeon Mac Pro in 2023? Are they just going to drop the Mac Pro or do they actually have a rivaling ARM processor?
Does it really matter for the Pro line, if there will be something that gives Xeon run for its money? Say twice as fast or twice as many cores for the same price?
Once Pro on ARM is out, and it probably will be the last in a lineup, there will be a lot of software with native ARM support. At that point (unless there is some INSANE advantage that Xeons would give), it is pretty much given that it will become ARM. Question is only in how quickly Apple can come up with something that is noticeable beating Xeons performance-wise.
71 comments
[ 4.1 ms ] story [ 143 ms ] threadFirst Arm-Based Macs to Be 13-Inch MacBook Pro and Redesigned iMac, Launches Coming in Late 2020 or Early 2021
For $150 you can buy a cheap Intel based PC laptop with Windows 10 on it.
I think Bootcamp might work for ARM based Linux and the ARM version of Windows 10. At least Raspberry PIs that are ARM based can run those.
Is there any point in a $150 computer, especially if it's $150 including Windows?
I read in another thread that volume is what propelled x86s popularity. Intel’s difficulty with 10nm has created an opening for other fabs and other architectures to catch up. If ARM can be shown to work as a desktop processor, why wouldn’t HP, Dell, Asus and the rest make ARM laptops?
Also other companies don’t have access to apple’s arm chips with great single core performance.
It's similar problem to how Chromebooks have still not really displaced Windows laptops except at the very low end of the market.
Although the significance of this seems lower now than at any time I can remember as more happens in a browser than ever. Windows 10 apparently is fully ARM compatible with a built-in x86 emulation layer, so perhaps it could work.
They already do, but they aren't as popular or as capable for two reasons:
The first is that only Apple can take advantage of any speed or power management improvements that a switch to a different CPU might provide, and that's because software development for Windows is very much not slaved to the App Store in terms of application development. That wasn't through lack of trying by Microsoft in the early 2010s (Metro/UWP and the disaster that was Windows 8/8.1), but as a consequence of that failure Windows applications aren't developed with a mobile-first paradigm or cross-platform support, so their ability to be managed by the OS is more limited.
The second (which is also an implication of the first) is that the only time ARM can actually outperform x86 designs is when a single-digit-watt power envelope is required. In Apple's case, it's required because they refuse to implement a proper thermal management solution in their laptops, and as such using processors that can reach 4 GHz for about 5 seconds until they hit 100C and have to aggressively throttle back are a waste (and their fans spin up to 100%, and excessively heat up the keyboard and the base of the laptop).
By going to a processor type that will ultimately be significantly slower but much more consistent in terms of performance, it's possible to both use alternate cooling solutions (like fully passive cooling) and provide a much more consistent user experience. And provided that the GPU is any better than Iris IGP, and that video encoding is otherwise accelerated (i.e. Quick Sync), professional (non-developer) users will still notice an upgrade.
And sure, while developers will lose in terms of raw performance, Apple knows that developers that go out of their way to use their products are more than willing to make compromises and concessions to use their ecosystem. On the other hand, those who are concerned about performance don't (and never) buy Macbooks.
Wouldn't surprise me if they also want to make the laptops even thinner once again.
It’s also speculated that with active cooling ARM will enable high performance Macs at a lower price point. It’s very unclear how well they will be able to match AMD and Intel at the highest performance levels.
https://support.apple.com/en-us/HT202043
https://news.ycombinator.com/item?id=23604060
https://news.ycombinator.com/item?id=23603852
https://news.ycombinator.com/item?id=23603866
https://news.ycombinator.com/item?id=23602434
A $1e12 company doesn't have to pick the "pinnacle" of ISA/CPU possible. They just have to pick one that is good enough to meet their financial goals.
Also, the kernel codebase, which is shared with iPhones, will be able to target a single architecture
Additionally, sharing the ISA means that the ARM reference cores will always be available as a backstop if Apple's custom chips start running into trouble for whatever reason. IIRC, Qualcomm was doing a lot of hopping between full custom cores and ARM reference ones depending on which one happened to work better some particular year.
It’s been speculated Apple could even license the x86 instruction set to include a x86 coprocessor in every one of their designs.
Apple's business is taking something that already is proven to work, and integrate it into something that makes money.
ARM is proven to work. It was designed for phones, tablets, routers, everything.
That means you could go to a factory like TSMC and they can reuse in excess of 95% of their hardware chip templates. You can reuse work done in iOS hardware too.
We are talking about freaking less than 10 nanometer scale designs!! The transistor's layout at those scales is not even square but calculated with quantum effects with very complex software, and retouched by hand with very expensive feedback using electron and X-rays microscopes.
That also means you can reuse the compilers that already exist, the transpilers and the toolchain in general.
Hardware designs need literally years of testing in very sophisticated machines(supercomputers) before you use them. Remember that if you make a mistake your entire die is wrong, million of dollars go to the toilet. And you always make mistakes before the design is mature.
It is not software that you could update easily and cheaply.
If you were to do your own thing , your chips would cost you 100x to 1000x to make just in money. It would also cost you orders of magnitude in time to market.
Of course a $trillion company can waste money away(until there is no more money left) but probably they want their money to give profit and create wealth instead of consuming it.
edit: I am aware not all Docker containers run on aarch64. But that has been changing already, and presumably would accelerate if Apple goes all in on ARM.
macOS is not linux, so it will need some form of emulation/virtualization to run "Docker on ARM macOS"
I do worry about software that is highly tuned with hand-rolled assembly (looking at you various video/photo/audio processing apps) and if this is going to be another transition that is mostly seamless, but with enough edge cases to leave people slightly bloodied.
So there's not going to be a Xeon Mac Pro in 2023? Are they just going to drop the Mac Pro or do they actually have a rivaling ARM processor?