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 Expected to Announce Arm-Based Mac Plans at WWDC, Transition Away From Intel to Begin in 2021
- Thread starter MacRumors
- Start date
- Sort by reaction score
Hello hailey bee,
Yep, my workflow does correlate with the benchmark.
I use Mathematica in many ways, utilizing its powerful symbolic, numerical and graphical capabilities in my calculations. In addition, some of the Mathematica programs that I've written are used to calculate stuff (have run-times of days)!
Currently, I wouldn't be able to use Mathematica on an ARM platform.
Of course if Apple releases ARM desktop computers which are equal (or surpass) my current 3.7 Ghz Quad-Core Intel Xeon E5 system (2013 Mac Pro), when running Mathematica, then I will be happy to support Apple with my money. Otherwise, no dice. Over time, my computing systems must go upward in capability, not downward.
richmlow
Yep, my workflow does correlate with the benchmark.
I use Mathematica in many ways, utilizing its powerful symbolic, numerical and graphical capabilities in my calculations. In addition, some of the Mathematica programs that I've written are used to calculate stuff (have run-times of days)!
Currently, I wouldn't be able to use Mathematica on an ARM platform.
Of course if Apple releases ARM desktop computers which are equal (or surpass) my current 3.7 Ghz Quad-Core Intel Xeon E5 system (2013 Mac Pro), when running Mathematica, then I will be happy to support Apple with my money. Otherwise, no dice. Over time, my computing systems must go upward in capability, not downward.
richmlow
Last edited:
I think the "arm v6" is a reference to the raspberry pi, and not something like the a12x. That said, wolfram wants to charge big bucks if you want to use more than 8 cores. If Wolfram continues to support the mac, will it also charge thousands more?
Last edited:
Hello jerwin,
It's a good point that you brought up, regarding the associated cost to use Mathematica (depending on # of cores)!
Yet another reason why I think I'll be hanging onto my Intel system for quite awhile, until all these ARM rumors are sorted out (fact from fiction)!
richmlow
Hi All,
On a side note....I keep hearing all this blather about A12X (and the like) ARM chips in iPads being so much faster than Intel chips, according to benchmarks, etc.
This is measured using a stripped-down OS (i.e., iOS and the like) on tasks which are "small".
Does anybody know the approximate performance of A12X (say), when forced to compute under a sustained heavy computational workload in real-life?
richmlow
It's a good point that you brought up, regarding the associated cost to use Mathematica (depending on # of cores)!
Yet another reason why I think I'll be hanging onto my Intel system for quite awhile, until all these ARM rumors are sorted out (fact from fiction)!
richmlow
[automerge]1591769279[/automerge]
Hi All,
On a side note....I keep hearing all this blather about A12X (and the like) ARM chips in iPads being so much faster than Intel chips, according to benchmarks, etc.
This is measured using a stripped-down OS (i.e., iOS and the like) on tasks which are "small".
Does anybody know the approximate performance of A12X (say), when forced to compute under a sustained heavy computational workload in real-life?
richmlow
Last edited:
On CPU bound tasks, even a relatively bloated OS will be roughly equivalent in speed during a benchmark run. Why? Benchmarks are typically run with high Cpu priority and background tasks are deferred / low priority. You're talking 1-5% difference, tops with a stripped down OS vs. something that isn't (e.g., iOS vs. MacOS). That isn't anything to write home about and won't significantly change the state of play.
RAM consumption of the OS is irrelevant to CPU benchmarks so much as there is enough to run the benchmark.
[automerge]1591769699[/automerge]
Sustained will be a matter of how well it is cooled. In a MacBook or desktop there will be a lot more space for a cooling solution than there is inside an iPad.
However, we can infer that due to the passively/almost not cooled iPads being competitive against lower end laptops running intel processors with active cooling, that given similar cooling and power consumption it is likely to be a bloodbath for intel so long as the software is native to ARM. If it's not, there will be an emulation/translation overhead and things become less clear.
Which is why apple are probably likely to include a just-in-time compiler/translator like rosetta - for existing apps that have not been updated. The app will likely be translated on the fly the first time you run it, and a version that's translated for ARM will likely be cached on disk after you've run it (or something similar).
Additionally, future apps will be "Fat" and include both x86 and ARM code, just like apple did in the PPC/x86 transition days.
Last edited:
in the end, it comes down to a translation-table between Intel and ARM processor instructions. This is very low-level stuff, that is not noticeable to applications so nothing will happen and virtualization will still work as we know now. I think we will see a new Rosetta stone that handles this translations for us, making OSX fully processor-agnostic for the end user. I assume nothing will change except machines will gain a great performance increase, less heat and better battery life. And systems will be much more modular like having a close integration with dedicated Security, GPU, AI, Augmented Reality processing chips. Of course, this is just an opinion and not a fact but some of it is already visible!
Last edited:
I’m waiting for Apple announcing the demise of macOS in favour of running iPadOS on “desktop” computer, and every professional user will instead buy an iPad Pro for all of their professional needs. Apple will do the exact thing they insist they will never do: merge touch screen interface and mouse keyboard interface together in their own way, like windows 10.
You might ask: what if I need more processing power? Apple will sell smart dock that houses more ram, more CPU, more GPU, much like Microsoft surface book. You dock your iPad on it, bam, more processing power transparently passed through iPad Pro and gets the job done. Take the iPad Pro out, and go about your business.
macOS was versatile because it exposed all of their Unix part of the system to advanced users and professionals to harness its true potential. With Apple locking down macOS getting rid of intel processors, I don’t see macOS as versatile as it was anymore.
You might ask: what if I need more processing power? Apple will sell smart dock that houses more ram, more CPU, more GPU, much like Microsoft surface book. You dock your iPad on it, bam, more processing power transparently passed through iPad Pro and gets the job done. Take the iPad Pro out, and go about your business.
macOS was versatile because it exposed all of their Unix part of the system to advanced users and professionals to harness its true potential. With Apple locking down macOS getting rid of intel processors, I don’t see macOS as versatile as it was anymore.
If Apple goes all in on ARM why would anyone want to purchase any Intel based Macs today ? Clearly Apple, and all developers, will be focusing on ARM based macOS as well as ARM based apps. Intel based Macs may be supported for a while longer but eventually will be left to die on the vine. Developers are not going to optimize code for two ecosystems. Apps will be written and optimized for ARM then virtualized or ported (not optimal) for Intel macOS. As already pointed out we would also lose out on running native or thin hypervisor based Linux and Windows.
It's somewhat exciting, but also concerning.
It's all well and good that Apple can transition their products from Intel to ARM, but just like with the last transition, thousands and thousands of apps will never be updated.
As someone who enjoys a game or two in the home, but also care about graphics rendering and HPC for work, Apple's hardware has been disappointing for a long time. The GPU's are always an afterthought. A switch to ARM might improve matters for me professionally if it leaves more room in the power budget for a higher end GPU, although HPC on ARM sounds... dubious at best.
And for gaming, well... if Apple switches to ARM, gaming on the Mac is over. I think it'll be mostly iOS ports. I don't see any chance whatsoever of Blizzard porting their games over as an example, and the Steam library will get wrecked all over again; and in this regard I think Valve will be fed up, too.
Also, I highly doubt that Windows will be supported. Yes, there is Windows for ARM, but ARM isn't unified like Intel. Windows supporting snapdragons does not mean it will support A14 or similar at all. Linux might work, but the T2 chip is going to make short work of that as far as installing it internally goes.
I understand the frustration with Intel these last few years, but I really do worry about software compatibility a lot on this. I guess we'll see where it all goes from here, but if this is true, I'm holding off buying any more macs yet again. The last one I bought was in 2011 - maybe it's time for me to just pack it up and give up. I just kept hoping and hoping that this time it'd be different. But it never was. I'm already on Linux full time...
iPhone's been great though, so there's that. But iPhone also keeps breaking my old apps, and it's frustrating sometimes.
It's all well and good that Apple can transition their products from Intel to ARM, but just like with the last transition, thousands and thousands of apps will never be updated.
As someone who enjoys a game or two in the home, but also care about graphics rendering and HPC for work, Apple's hardware has been disappointing for a long time. The GPU's are always an afterthought. A switch to ARM might improve matters for me professionally if it leaves more room in the power budget for a higher end GPU, although HPC on ARM sounds... dubious at best.
And for gaming, well... if Apple switches to ARM, gaming on the Mac is over. I think it'll be mostly iOS ports. I don't see any chance whatsoever of Blizzard porting their games over as an example, and the Steam library will get wrecked all over again; and in this regard I think Valve will be fed up, too.
Also, I highly doubt that Windows will be supported. Yes, there is Windows for ARM, but ARM isn't unified like Intel. Windows supporting snapdragons does not mean it will support A14 or similar at all. Linux might work, but the T2 chip is going to make short work of that as far as installing it internally goes.
I understand the frustration with Intel these last few years, but I really do worry about software compatibility a lot on this. I guess we'll see where it all goes from here, but if this is true, I'm holding off buying any more macs yet again. The last one I bought was in 2011 - maybe it's time for me to just pack it up and give up. I just kept hoping and hoping that this time it'd be different. But it never was. I'm already on Linux full time...
iPhone's been great though, so there's that. But iPhone also keeps breaking my old apps, and it's frustrating sometimes.
In 2026, a 2019 MacPro would probably be due for a processor upgrade. The Mac Pro is designed to be highly modular and, according to iFixit (see below), that moduarity should extend to the processors, which appear to be upgradable. So if Apple honors its machine's capability, at that time one should be able to upgrade to the contemporaneous ARM processor / motherboard / I/O, and make the machine current again:
Mac Pro 2019 Teardown
The Mac is back and more Pro than ever, throwing away the cylindrical “trash can” design in favor of something that resembles a computer. Its...
www.ifixit.com
A: We haven’t tested them yet, but to all appearances: Yes to processor upgrades, and yes to RAM upgrades. Both are modular, socketed, standard components.
Jeff Suovanen - 12/17/2019
Last edited:
Let's hope they still have Rob on their team...
To be honest, I'm not following why people are so quickly making assumption (and judgement) on this and using the Mac Pro as the reason.The Mac Pro is designed to be highly modular and, according to iFixit (see below), that moduarity should extend to the processors, which appear to be upgradable. In five years, a 2019 MacPro would probably be due for a processor upgrade. So if Apple honors its machine's capability, at that time one should be able to upgrade to the contemporaneous ARM processor / motherboard / I/O, and make the machine current again:
Q: Is it possible to upgrade processor after buying the computer?
Mac Pro 2019 Teardown
The Mac is back and more Pro than ever, throwing away the cylindrical “trash can” design in favor of something that resembles a computer. Its...
A: We haven’t tested them yet, but to all appearances: Yes to processor upgrades, and yes to RAM upgrades. Both are modular, socketed, standard components.
Jeff Suovanen - 12/17/2019
With the intel transition, Apple started with with portables to follow with their narrative of performance per watt. If the ARM transition is true, I would bet that they will start off with the consumer models of portables, or the Mac mini, with heat/power throttling being the reason. The Mac Pro is probably going to be the last, if ever, product to be transitioned.
Heck, for all we know, Apple could simply start a new line of products and keep the Mac as is, to avoid confusion and expectation. We'll see what happen comes WWDC.
[automerge]1591774195[/automerge]
If Johnny Ive was still around, that concern is completely valid.
However, we already see Apple moving away from design over function. iPhones are getting thicker/heavier and larger batteries, butterfly keyboard regressing back to more reliable scissor mechanism, and then there's the highly modular (for a Mac) Mac Pro.
The cooling part is probably why Apple is pushing with ARM transition.
Last edited:
The possibility of exchanging components (repair, replacement, extension) will not improve.
But I agree that some things have improved (especially the keyboard).
Cooling:
Apple generally makes mistakes here and tends to design the cooling too poorly. This has nothing to do with ISA. Even with ARM, Apple can make questionable cooling and design decisions (e.g. for aesthetic).
The ARM advantage in terms of power consumption and cooling is, in my opinion, totally overestimated by many. ARM is not a miracle. There is of course the x86-overhead. But in relation to the total die area of a complete CPU (transistor count) it is rather small (don't forget big caches, I/O, IGP/GPU etc.). ARM has some advantages here, but depending on the CPU (or SoC), they are not as big as many people think. A very powerful ARM-CPU will also be consuming some power...
How good a CPU is depends not only on the ISA, but also on the talent of the design team, the chip design and the built-in features. Of course the manufacturing process also plays a big role.
Of course Apple can build in exactly the desired features with an own CPU.
The switch to ARM may be correct.
But the situation today is also completely different than it was with the Intel switch.
Last edited:
Do you know what the vintage of that ARM processor is? It could be quite old. Note they've even got a G5 in there. I also use Mathematica extensively, and understand that Wolfram has been working to optimize Mathematica for the ARM architecture.
At this point we really don't know how the performance of Apple's ARM processors will compare with Intel's for key programs like Mathematica, MS Office, Adobe Photoshop etc. We'll just need to wait for product release and independent real-world benchmarks.
Porting is certainly a potential issue, as we saw with Microsoft's ARM-based Surface Pro X:
Last edited:
Of course. Personally I believe the nudge to switch lies more in GPU power (in addition to performance per watt). Apple has been pushing hard on GPU in the past years, and the Macs are bottlenecked by intel's integrated graphics (and Apple's dislike of nVidia). The fact they are pushing the GPU in the Ax chip so hard shows that they really are obsessed with GPU power.
It should be noted that there is at least one high-powered ARM chip currently in commercial use: the Fujitsu A64FX, which has been incorporated into a few new supercomputers:
www.networkworld.com
However, the significant efficiency advantage that ARM chips are supposed to have over those from Intel at the low-power end doesn't necessarily scale to the high-power regime. If you look at the latest Green 500 list (Nov. 2019), which ranks supercomputers by power efficiency (GFlops/watts), systems based on the A64FX, Intel Xeon, and IBM POWER 9 can all be found in the top 10.
What I've picked up from my general reading is that, for HPC, the ARM, Xeon, and AMD processors each have their own strengths and weaknesses, and which will be superior depends on what tasks you need them to perform.
The only consistent advantage that the A64FX currently demonstrates is lower cost. And of course, when it comes to its own chips, Apple would love to have lower cost — and direct control. Whether Apple can additionally beat Intel at its own game in the high-performance regime remains to be seen. If it can, that would be icing on the cake.
Cray to license Fujitsu Arm processor for supercomputers
HPE's Cray will co-develop Fujitsu's A64FX CPU to meet the requirements of likely customers such as universities and national research laboratories.
Breaking down the memory-speed bottleneck
This spring, Labs researchers anticipate Sandia becoming one of the first DOE laboratories to receive the newest A64FX Fujitsu processor, a Japanese Arm-based processor optimized for high-performance computing.
www.sandia.gov
However, the significant efficiency advantage that ARM chips are supposed to have over those from Intel at the low-power end doesn't necessarily scale to the high-power regime. If you look at the latest Green 500 list (Nov. 2019), which ranks supercomputers by power efficiency (GFlops/watts), systems based on the A64FX, Intel Xeon, and IBM POWER 9 can all be found in the top 10.
What I've picked up from my general reading is that, for HPC, the ARM, Xeon, and AMD processors each have their own strengths and weaknesses, and which will be superior depends on what tasks you need them to perform.
The only consistent advantage that the A64FX currently demonstrates is lower cost. And of course, when it comes to its own chips, Apple would love to have lower cost — and direct control. Whether Apple can additionally beat Intel at its own game in the high-performance regime remains to be seen. If it can, that would be icing on the cake.
Last edited:
Tbh, I can see why they're doing this - but there has been such a big push from AMD in the x86 space, I would question whether Intel -> AMD would be a far more appealing jump than Intel -> Arm (except maybe from a profitability standpoint). And Apple already have a good commercial relationship with AMD in the graphics department.
I mean, Apple have resorted for a while now to using server grade CPUs in their pro desktop lines, with AMD that wouldn't be necessary. Imagine a Mac Pro with a 3990x! God I'd buy one of those in a heartbeat.
I mean, Apple have resorted for a while now to using server grade CPUs in their pro desktop lines, with AMD that wouldn't be necessary. Imagine a Mac Pro with a 3990x! God I'd buy one of those in a heartbeat.
I am sorry techhwiz , I read too much misinformation in this thread ,and as a long time lurker I find your confidence and '"you can argue but i design chips" a bit of too much condescending for me to let go , lets start by the fact that you are incorrect , with all due respect to PCIE switch design saying that makes you "designing chips for a living you cant argue with me" is a big nono in our business (I am like you "design chips for a living" for the last 15 years).
First your Cortex-M comment is correct but out of context , ARM provides a custom instruction module to integrate (optional) that supports a predefined "custom" instructions set that you can implement this is of course a non optimized module that helps accelerate DSP like process (you can add the DSP extension as well if you like) but the new custom instruction are not part of your "normal" pipeline and due to this complication (I guess this part) of integrating new "custom" instruction set into a complicated OOO pipeline that needs to support exceptions with restart and the likes make its an M only feature.
If you understand what an instruction mean , how it is implemented , how it is being used by the compiler and finally SW you will understand your fallacy.
Quick example from the X86 world to make it clear , we both agree that both Intel and AMD (and VIA
) support X86-64 Uarch , BUT Intel has "private" extensions , which due to legal agreement they are forced to cross license for free to AMD and VIA , so for example when they introduced the AVX-512 instruction set as an extension they were the only ones supporting this in the compilers and SW that supported it , i.e GCC 4.9 could use the new instructions or Blender for example could do so as well from a certain point.In no point in time Intel stopped being X86-64 compliance even though they did add a new instruction that is not used by the rest of the X86-64 designs , until this day we dont have AMD supporting native AVX-512 instructions.
TLDR - You are wrong about your claim about instruction set usage in modern CPU`s , but unlike you i will be happy to discuss this further! you are welcomed to argue if you like
.No, it's not as fast as the i7-9750h, but it's Geekbench score comes quite close and even exceeds it on single core:
I totally concede we need to take that with a large pinch of salt seeing as it is just a Geekbench score, but there are real world videos on Youtube showing how fast the iPad can handle things like exporting video compared to traditional laptops.
Apple claims the iPad Pro is faster than 92% of laptops sold, which is a pretty big caveat, but I'd be excited to see what the chip could do inside a thicker form factor with better cooling and increased power consumption.
The performance is irrelevant if you cant trust the environment you are coding on to be the same as where you deploy. Yes I’ll be able to continue my hobby on macOS and photo editing might get faster. But for programming that is completely a no go.
The export is done by hardware and not by the CPU itself. Fixed-function hardware vs. general-purpose CPU. Therefore it is not a good comparision.
Fixed-function hardware is very fast but also inflexible. The encoding quality is usually worse than very good software encoders and improvements are not possible due to lack of updates.
I just don't see how is this possible. Not one computer out there succeeded on the ARM platform. All the software in the world works on Intel, and I just don't see how Apple is going to convince app developers to rebuild their software to work on ARM on a diminishing computer market for the small market-shared Mac OS, which within it they will only target brand new buyers and forget about all the installed intel user base.
My only guess is Tim betting big on the iOS software library, they might as well combine the iPad Pro with the MacBook and call it the "PadBook".
From 2020, Yes Mr.Jobs, you were right, indeed it is. And Windows still suck.
My only guess is Tim betting big on the iOS software library, they might as well combine the iPad Pro with the MacBook and call it the "PadBook".
From 2020, Yes Mr.Jobs, you were right, indeed it is. And Windows still suck.
More likely, a few years down the line, processor architecture will be more diverse because any software that isn't actually an operating system kernel or a just-in-time script compiler will be architecture independent, and written in scripting languages or compiled to bytecode.
That's not so far fetched: modern Windows applications written in C# etc. are CPU-independent bytecode (it's the old legacy apps that are a problem). Android apps are mostly bytecode. Apple's XCode supports compiling to bytecode which (if I understand correctly) the App Store translates to the appropriate processor when the App is installed. GPU-based computing and other acceleration devices are best implemented as operating-system level frameworks (like Metal and Accelerate) so that existing software can take advantage of new acceleration technologies. (Those ARM supercomputers get most of their "super" by filling the space freed up by smaller ARM cores with shaders and vector processors)
What we're really seeing is the end of a "glitch" caused by the IBM PC locking the market intro a single processor (...and one which started as a kludgey stop-gap, at that). That was a regressive step even back in the 80s - Unix pre-dated the IBM PC by a decade and was always focussed on source-level compatibility, but just as cheap microprocssor-based personal computers were leaving the 8-bit era and getting powerful enough to run Unix-like operating systems and not rely on lovingly hand-crafted assembler, along come IBM and Microsoft and shackle the industry to an outdated CP/M knockoff running on a slightly souped-up version of the 8-bit 8080....
...apart from the iPhone, the iPad and all those Android devices which have completely changed the face of modern computing...
Apple are now in the position that the iPad Pro is more powerful than the Air, and any 12" MacBook replacement they could build with Intel... and that's still being healthily skeptical about the benchmarks that show it competing with the 15" i7 MBP (maybe not, but it's still going to run rings around the Air).
With every passing year, application software gets less and less dependent on the processor architecture - it's really just the corporate sector who are hooked on ancient binary software, and Apple have never been strong in that sector.
Good post.
I don't think Apple will be far out with the laptops and 90% claim. £299 crap tops are iPad cannon fodder.
Those geekbenches are quite dramatic. And a fanless cpu/gpu.
And if Geek bench won't suffice. We shouldn't over look the fact that Affinity apps have come to iPad with most of the features intact and with great performance. Or that Procreate on an iPad is blistering compared to Painter 2019 on a Mac. Procreate is what you get when you have a 1st rate developer on your hallowed ground. Vs a journeyman developer giving you a back handed port.
Or deprecated open gl with deprecated gaming ports.
System on a chip and integration is likely the future as devices get more compact and svelte. That's probably why Intel are going to get left behind, as Apple engineers lack confidence in their future road map. Just as they ran out of patience with the IBM G5 road map.
If Apple can't make the devices they want to...they'll move on.
It just so happens they've cultivated their own CPUs. Rejoice!
The proof is in the pudding though. What devices? New market defining ones? Or slimmer versions of our current 'Macs' with just as good performance...or better? We won't know until they're revealed. But I don't see Apple doing this to give us 'Atom' (by Intel) style performance. It has to be a compelling performance argument.
I'll be following the progress for the Mac ARM chip with great interest.
Azrael.
[automerge]1591792126[/automerge]
Well, for prices? I hear what you're saying. All those who gain power are frightened to lose it.
And Apple have ridden prices up under the guise of 'Pro.' Sometimes by £50, £100, Several hundred. £435. £3000. On product lines. To me. Outrageous.
iOS, just because they're on ARM, hasn't seen made much difference to prices on new kit.
So for Mac ARM. We may get prices rises and Apple pocketing the costs they give to Intel.
Azrael.
[automerge]1591792481[/automerge]
Azrael.
Last edited: